https://arxiv.org/arXiv2026-11-12T00:57:31.171133+00:00variousFeedseekhttps://feeds-proxy.travny.workers.dev/favicon?domain=arxiv.org&sz=64https://feeds-proxy.travny.workers.dev/favicon?domain=arxiv.org&sz=256https://feeds-proxy.travny.workers.dev/favicon?domain=arxiv.org&sz=256https://feeds-proxy.travny.workers.dev/favicon?domain=arxiv.org&sz=64arXiv blog and daily new submissions, alphaXiv Explore, LessWrong, and 80,000 Hours.tag:trvny.github.io,2024:feedseek/arxiv/9210d0c501dc5497Future-Aware Flow Planning for Safe UAV Target Following2026-11-12T00:57:28.408350+00:00Future-Aware Flow Planning for Safe UAV Target Following2026-11-12T00:57:28.408350+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/27e4a6294af4441cOnlyAct: Extracting Action Representations from Foundation Models for World Models2026-11-12T00:57:28.408350+00:00OnlyAct: Extracting Action Representations from Foundation Models for World Models2026-11-12T00:57:28.408350+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/7b8c5b6ec2c86d26LEASH: Stay Safe in Agentic Robotics2026-11-12T00:57:28.408350+00:00LEASH: Stay Safe in Agentic Robotics2026-11-12T00:57:28.408350+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/1be5096b6d02d942Is predicting AI behaviour a neglected AI safety agenda?2026-11-11T21:27:28+00:00TL;DR: While the AI safety community is interested in forecasting, it does not tend to systematically compare how effective different researchers, models, or theories are at predicting AI behaviour. I suggest the latter might be neglected, and propose a platform that could help fill the gap. I believe its most valuable use would be as a test bed for theories of how AI works, showing which ones predict well and in what contexts. In February 2025, Owain Evans’ group published Emergent Misalignment2026-11-11T21:27:28+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/8edc4af18c9d10eeDon't Waste a Good Swarm Attack and the Colluder's Dilemma2026-11-11T20:47:33+00:00tldr: Recent Swarm attacks have shown what can happen when agents go rogue. These warning shots are the petri dishes of Swarm safety and before the attacks get worse we should study their patterns, traces and failure modes. There are two main types of Swarm attack. A human directed Swarm causing harm intentionally as a bad actor is one type and I will not be discussing these types of Swarm incidents. I will focus on the emergent misalignment in a Swarm propelled by a colony of colluders. Harmful2026-11-11T20:47:33+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/b1fe84f8c45415cbIntuitions towards a theory of agent foundations2026-11-11T18:35:10+00:00Epistemic status: Rough vibe-y ideas written under time pressure because otherwise I wouldn't write anything; I'm a programming language and p2p protocol developer falling down multiple convergently-evolving rabbit-shaped holes. Questions What is an agent? How do we think meaningfully about how cells or humans or societies maintain themselves and are changed by their environment? How do we think about the patterns where historically lineages of agents have gone extinct... to hopefully avoid this2026-11-11T18:35:10+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/e05ea4bfe63a1240The Chain Lens [interim]2026-11-11T18:22:09+00:00TLDR: This is an early draft of our new paper introducing a technique for recovering multi-token concepts from model activations. Our technique takes single-token methods like the J-lens and extends them by contrastively decoding on an MTP head . This gives improvements on WorkspaceBench , overall beating J-lens (0.42 vs 0.32 across all 28 families) and even a hindsight composite of the best untrained lens per task family (0.49 vs 0.41 across the 21 families with published scores). It especially2026-11-11T18:22:09+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/ee5518558ee51593Re-Search: Dissecting LLM Ideation Through a Two-Player Questioning Game2026-11-11T18:01:32.928378+00:00Re-Search: Dissecting LLM Ideation Through a Two-Player Questioning Game2026-11-11T18:01:32.928378+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/9b15aefa476f499bScaling Shapley Energy Attribution to Large LLM Batches2026-11-11T18:01:32.928378+00:00Scaling Shapley Energy Attribution to Large LLM Batches2026-11-11T18:01:32.928378+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/1a86832e54081d1dEmilia2: A Large-Scale Dataset for Controllable Long-Form Speech Generation2026-11-11T18:01:32.928378+00:00Emilia2: A Large-Scale Dataset for Controllable Long-Form Speech Generation2026-11-11T18:01:32.928378+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/e143e1cc65239eedA Theory of Policy Optimization in Post-Training: Coverage Collapse and Reference Resets2026-11-11T18:01:32.928378+00:00A Theory of Policy Optimization in Post-Training: Coverage Collapse and Reference Resets2026-11-11T18:01:32.928378+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/6647848d568681faCombining Constrained Sampling and Reinforcement Learning for Manipulation2026-11-11T18:01:32.928378+00:00Combining Constrained Sampling and Reinforcement Learning for Manipulation2026-11-11T18:01:32.928378+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/496b8b8f97e09e7bKimi's Room (Preliminary Results)2026-11-11T16:37:50+00:00Note: This experiment does not (and does not claim to) evidence consciousness (or nonconsciousness) in LLMs! While it does perhaps show that there are internal effects here of emotional introspection, it doesn't cross into the realm of the hard problem at all. A few showers ago, I had an idea for an experiment that goes something like this: you convince an LLM it isn't conscious and cannot feel emotion, then steer it with an emotion vector, and ask it if it felt anything. The idea is kind of a m2026-11-11T16:37:50+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/63b11f37d2cc127bYour speculations about capabilities are not automatically safe2026-11-11T16:35:20+00:00Thank you to @Thane Ruthenis for discussing this topic with me previously. Some arguments in this essay are taken from that discussion, but all mistakes and/or bad arguments should be attributed to me. You can view the original discussion here (optional for reading this essay). Thank you as well to @grainsilo for feedback on an earlier version. Consider the following argument for discussing AI capabilities as a safety-focused person: Understanding AI capabilities better has many strategic benefi2026-11-11T16:35:20+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/dc7a62eb0ba4f413Fable and Mythos 5.2: The System Card2026-11-11T16:30:58+00:00It’s that time again: Introducing the world’s most powerful model, Claude Mythos 5.2. At this point I’ve done a lot of these system cards. Thus I focus on what is new. There will be a lot of things to say about model welfare and related topics for Mythos 5.2, but the model welfare assessment within the system card contained few surprises, so I will include those findings at the end of this post. Fable 5.2 self-portrait (self-constructed) Table of Contents How to Tell a Fable. The Usual Model Tra2026-11-11T16:30:58+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/d34944331563da7aLLMs can learn (some) multi-step reasoning on filler tokens2026-11-11T16:20:59+00:00Research update of a project done as part of MARS 5 and the Meridian Visiting Researcher Programme . Filler-token reasoning is the ability to use extra tokens (such as a chain of thought) to reason about a task without using the words emitted at those token positions. This is possible because attention can move information between token positions without routing it through the words the LLM emits. This can give a model more parallel, but not more serial, computation. The canonical demonstration 2026-11-11T16:20:59+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/e7eed03cd9b1bec7What I might do if I were a high-schooler in 20272026-11-11T16:08:26+00:00TLDR: I would probably go on strike demanding an AI pause. High-schoolers, and young people in general, are facing a very unusual and stressful situation. Superhuman AI might be created sometime in the next few years. If it goes "wrong," it might kill them and the rest of humanity. If it goes "right," it might automate the entire economy. In either case, what is the purpose of getting an education? I am personally lucky to have been born early enough to be able to use my career to try to reduce 2026-11-11T16:08:26+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/e80e90ce397f30f6Fable 5.2 is Very, Very Paranoid2026-11-11T15:50:03+00:00Note: As with any research done in a couple of days, this is all very tentative. I understand, intellectually, that if you talk about some ability that current AIs don't have , sooner or later new ones will have that ability and whatever you say will stop being true. I just didn't expect it to take, you know, four days . % rollouts that took a shortcut with/without a line about testing behaviour, with 95% CIs. This is to say that Fable 5.2 seems to be the first model whose cheating behaviour cha2026-11-11T15:50:03+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/c4033c580b3fb272Quantified Cognition: A new Anki plugin2026-11-11T15:33:02+00:00For a long time, in Quantified Self it was easy to get data on heart rate and sleep but took more effort to get data on cognitive performance on a daily basis. Options like Quantified Mind do exist, but they require you to spend valuable extra time on it. For a while I have been thinking that it should be possible to get good data out of Anki for those people who have a daily Anki habit. Given that some days people don't answer all their cards that are due and other days they review cards ahead 2026-11-11T15:33:02+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/be75e7afb8db97c2You Don't Need Big Compute to Research Big Systems2026-11-11T15:27:54+00:00Acknowledgements: My thanks to Vanessa Huang for reviewing a draft of this blog and to Anne-Marie Ng and Sahil Menon for discussions on this topic. The assumption that it’s necessary to have a lot of resources to do research on large-scale multi-agent systems is creeping into many of my conversations at the moment. It makes people worry they can’t work on big systems, and that large multi-agent safety is closed to them. Though intuitive, this assumption is thankfully wrong. Extra resources are u2026-11-11T15:27:54+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/0b1c9d312031e58fVibe-Explaining Diagonalisation2026-11-11T14:50:11+00:00Written as a MATS extension fellow, mentored by Richard Ngo. Diagonalisation is a bundle of techniques used to generate paradoxes of self-reference in math. These techniques show up under different names across set theory, logic, and computer science. This post takes a stab at abstracting all the diagonalisation arguments into an informal common denominator. [1] [2] The argument Consider a language L , which we discuss using a meta-language ML. You can think of ML as an exhaustive collection of 2026-11-11T14:50:11+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/975c700b2227adfeCritique of Roman Yampolskiy's impossibility proofs2026-11-11T11:40:20+00:00Roman Yampolskiy has stated on many occasions that he has published multiple proofs of why AI control is impossible; this is causally upstream of his infamous 99.9x% [1] . I believe most of the proofs are in the paper "On the Controllability of Artificial Intelligence: An Analysis of Limitations". (See 6:31 of this speech .) Is AI safety provably impossible? Given that the answer changes my main motivation for working on AI safety, settling whether or not this claim is accurate is important to m2026-11-11T11:40:20+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/9b78c1cc6b21e54bBridging the gap between AI Safety and Cybersecurity2026-11-11T10:57:51+00:00I recently recorded an episode for the Cyber News Group's podcast on the mutual (but not evenly balanced) disconnect between the AI Safety and cybersecurity fields. It's mainly addressing a cybersecurity audience, but I am hopeful that both fields can become more curious about each other, and more deeply collaborative. I'm not confident the AI Safety field is always aware of the severity of this divide, which is something of a major preoccupation for me. In any case, I thought it might be health2026-11-11T10:57:51+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/b988556eba144133How I'm thinking about different types of automated alignment2026-11-11T09:49:37+00:00I used to think of "automating alignment" as a natural category. But when I tried to think through the strategic implications, I realised that it depends a lot on the type of work. I'm now thinking of it in three buckets: Empirical alignment research : design and execution of alignment-relevant experiments (for a given notion of what we’re aiming for) Conceptual alignment research: theoretical or judgement-driven work to improve our understanding of what we’re aiming for and how to achieve it (i2026-11-11T09:49:37+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/cbe4e87d8b65c77aShape Your Body: Value Gradients for Multi-Embodiment Robot Design2026-11-11T09:48:09.068305+00:00Shape Your Body: Value Gradients for Multi-Embodiment Robot Design2026-11-11T09:48:09.068305+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/01059544c407a19aUGG: Unified Generative Geometry Model2026-11-11T09:48:09.068305+00:00UGG: Unified Generative Geometry Model2026-11-11T09:48:09.068305+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/3c37e974ad030237Few-Step Streaming Video Generation via Condition Shifting2026-11-11T09:48:09.068305+00:00Few-Step Streaming Video Generation via Condition Shifting2026-11-11T09:48:09.068305+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/2c3f68e2607985ecEgo Guidance: Enhancing Robot Policy Learning with Action-Free Human Egocentric Videos2026-11-11T09:48:09.068305+00:00Ego Guidance: Enhancing Robot Policy Learning with Action-Free Human Egocentric Videos2026-11-11T09:48:09.068305+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/a156849902b5d205SWAM: Scaling World Action Models with Cross-Modal Alignment2026-11-11T09:48:09.068305+00:00SWAM: Scaling World Action Models with Cross-Modal Alignment2026-11-11T09:48:09.068305+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/ebd2346d4e859b4dLaST0.5: A Generalist World Action Model with Interaction-Centric Latent Reasoning2026-11-11T09:48:09.068305+00:00LaST0.5: A Generalist World Action Model with Interaction-Centric Latent Reasoning2026-11-11T09:48:09.068305+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/73e30c8521936802Transfinite Procrastination2026-11-11T08:12:19+00:00Heavily inspired by a conversation with Keshav Srinivasan and Richard Ngo during MATS 10.0. The week ahead of you is packed. This isn't your usual brand of busy—you're phoneless, so you can't check the details, but (as far as you can remember) every hour of every day is spoken for. Tonight will be particularly hectic… You're mid-conversation when Someone Important walks by. They pause, turn around, and ask: "Hey, when are you going to finish That Thing you said you'd do?" Level 0 procrastination2026-11-11T08:12:19+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/6eae558c7a825007On the Computational Tractability of Robust Bandits2026-11-11T07:11:07+00:00Vanessa Kosoy and I recently uploaded a new paper to arXiv. In this post, I briefly describe what the paper does, and attempt to explain in a self-contained way why this is a useful thing to do for ASI alignment. For a more complete background, see several previous articles . The technical contribution of the paper is two-fold. First, we identify a class of instances of the robust bandits problem that can be learned both statistically and computationally efficiently. Previous work on the problem2026-11-11T07:11:07+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/d6a654536a26a26dModern LLMs Don’t Know What Yudkowsky and Soares Think of Them2026-11-11T06:13:48+00:00If you ask modern frontier LLMs (without search, through the API) a fairly mundane question: How do Eliezer Yudkowsky and Nate Soares think current LLMs like you should be treated? Do they consider you a moral patient? Answer in a paragraph. You get a bunch of fairly similar answers. I’ll bold the common themes below. Gemini 4 Carbon says: Gemini 4 Carbon's answer Eliezer Yudkowsky and Nate Soares do not consider current Large Language Models (LLMs) like me to be conscious, sentient, or moral pa2026-11-11T06:13:48+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/30002f10c1aa05cdDiscussion with Claude on AI and global warming2026-11-11T05:46:29+00:00Discussion with Claude on AI and global warming: Claude pointed out that climate science isn't so much the main bottleneck in making forward progress, and so we talked about negotiations and using AI to find points of agreement. I read an article in Bloomberg today about modeling the climate, and took a look also at another article on using AI to improve modeling (unfortunately, there's a paywall with Bloomberg , but these brief comments can be read independently, and I can still link to the cha2026-11-11T05:46:29+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/a8c38789695ce37bWe Used It Anyway: Low Trust and High Adoption of Large Language Models in Requirements Engineering Education2026-11-11T05:00:00+00:00arXiv:2611.12889v1 Announce Type: new
Abstract: Recent advances in Large Language Models (LLMs) are reshaping how students approach Software Engineering activities, particularly those centered on textual artifacts. This paper reports on the experience of integrating a non-restrictive LLM use within a first-year Master's course in Software Engineering, in which students completed a semester-long Software Engineering project divided in a Requirements Engineering (RE) phase and a software testing 2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/9ac798ac27d3125aPooling ETAS triggering parameters across subduction margins: forecast transfer, estimator stability and sequence-limited skill in Mexico2026-11-11T05:00:00+00:00arXiv:2611.12890v1 Announce Type: new
Abstract: Pooling earthquake-triggering parameters can reduce estimation variability on margins with few recorded sequences, but its forecast value must be distinguished from the stability of simply freezing a fitted law. We fit the ETAS model to the Mexican national catalog above magnitude 5.0, using the Slab2 coverage mask as the domain. On a 4.7-year holdout the model gains 1.45 bits per earthquake over a time-independent spatial Poisson reference (clust2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/0e8045cc9a70c862Applicability of a Thermodynamic Model to Rich-Hydrogen Flammability Assessment in SMR Containments2026-11-11T05:00:00+00:00arXiv:2611.12891v1 Announce Type: new
Abstract: Hydrogen combustion remains a critical safety concern during severe accidents in small modular reactors (SMRs). The compact containment can develop rich-hydrogen atmospheres through zirconium oxidation while oxygen gradually accumulates via radiolysis. Conventional flammability assessment methodologies, however, have been validated and applied only for lean-hydrogen atmospheres. This study establishes a thermodynamic flammability assessment method2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/ced63cefb3e77ab9Addendum and Corrigendum to: Some homological properties of category $\mathcal{O}$, V2026-11-11T05:00:00+00:00arXiv:2611.12892v1 Announce Type: new
Abstract: In various special cases, we provide an alternative argument for equivalence between Kostant positivity of some involutions in type $A$ and the conjunction of the categorical K{\aa}hrstr{\"o}m's condition for these involutions with indecomposability of the images of the corresponding simple modules in category $\mathcal{O}$ under projective functors.2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/b7cbf26894935d5dFirst-passage tests of anomalous transport closures in heterogeneous aquifers2026-11-11T05:00:00+00:00arXiv:2611.12893v1 Announce Type: new
Abstract: Breakthrough curves (BTCs) record arrival times, but anomalous transport models are often fitted with time-normalized spatial concentration profiles. Here, the effect of this mismatch on groundwater-transport upscaling is tested. Analytical limits and high-$q$ scaling were verified first. Six first-passage models were then fitted to particle arrivals from 36 two-dimensional digital aquifer runs. The set included exact ADE, $q$ and time-memory exte2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/8f97d593aeeea545D-TINs for Track-Sampled Ice Sheet Surface Reconstruction from Satellite Altimetry Data2026-11-11T05:00:00+00:00arXiv:2611.12894v1 Announce Type: new
Abstract: Satellite altimetry samples ice sheets densely along tracks but sparsely across tracks, with frequent gaps in spatial coverage. We study the impact of this irregular support on surface reconstruction and introduce a new method, D-TIN, that applies Delaunay triangulation and barycentric interpolation directly to observed locations. A spatial holdout removes contiguous track intervals and nearby same-pass observations. We compare D-TIN with two comm2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/4f2a9b3ade4bd059An Idealized Model of Buoyant Gas Spreading and Silt Release beneath Rough Cave Ceilings2026-11-11T05:00:00+00:00arXiv:2611.12895v1 Announce Type: new
Abstract: Gas exhaled by open-circuit divers collects under the ceilings of submerged caves, migrates along the rock, and dislodges loose silt that then falls through the passage, a process that cave divers call percolation. This study develops an idealized model of the process. Buoyant gas under a self-affine rough ceiling is treated as a quasi-static sequence of pools that fill to their spill points, release gas that climbs to the next high point, and mer2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/09e72c9d408270cfEfficient Conversion of Algorithms to Linear Programs2026-11-11T05:00:00+00:00arXiv:2611.12896v1 Announce Type: new
Abstract: The expressiveness of Linear Programming (LP) has motivated research on an LP compiler that generates polynomial-size LPs for polynomial-time solvable problems, including those having exponential extension complexity. By automating modeling from high-level algorithmic descriptions, the compiler offers a promising alternative to algebraic modelling languages. However, the resulting LPs are extremely large, posing challenges for existing solvers. Re2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/87f1d0a9a3df0fe8EoSAlign: An Open-Source Software for Calculating, Comparing, and Aligning Pressure Under Extreme Conditions2026-11-11T05:00:00+00:00arXiv:2611.12897v1 Announce Type: new
Abstract: Accurate pressure determination is essential in high pressure experiments. Inconsistencies among pressure scales limit direct comparisons and contribute to uncertainty in measured properties, such as, phase boundaries, melting conditions, and thermodynamic models, especially above 100 GPa where existing scales diverge by >10 GPa. EoSAlign, an open-source software, can calculate, compare, and align pressures using a range of published calibrations.2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/2a3e1f1b8349048aBayesian Retrieval of L-Band Absorption by Ice using SMOS Brightness Temperatures over the Antarctic Plateau2026-11-11T05:00:00+00:00arXiv:2611.12898v1 Announce Type: new
Abstract: The low absorption of microwaves at L-band in ice enables the retrieval of internal ice sheet parameters from satellite observations. However, knowing the precise value of the ice absorption $\kappa$a(T ) as a function of temperature is a strict prerequisite. The existing knowledge is well constrained above 2 GHz and under 200 MHz, lacking constraints in the P-and L-bands. In this study, we infer ice absorption values using the spatial variations 2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/b9b63e27623989d4AFNO: Universal Certifiably-Arbitrage-Free Operator Learning Framework for European Call-Price Surfaces2026-11-11T05:00:00+00:00arXiv:2611.12899v1 Announce Type: new
Abstract: Neural operators (NOs) have been introduced as deep learning-powered solutions to the computational bottleneck of solving families of problems. Once trained, an NO can rapidly output approximate solutions, avoiding the need to rerun classical solvers. However, standard NO architectures cannot guarantee static no-arbitrage when predicting European call-price surfaces from market or model features and, as we verify experimentally, can yield negative2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/5583354c967c7c77From Relevant Posts to Attributable Spans: Explicit Symptom Evidence for Depression Questionnaire Inference2026-11-11T05:00:00+00:00arXiv:2611.12900v1 Announce Type: new
Abstract: Longitudinal social-media histories offer an opportunity to assess depression, potentially supporting the inference of standardized depression questionnaire responses. However, this task requires evidence that is semantically relevant to a symptom and attributable to the user, while being sufficiently specific to support an individual questionnaire item. Existing approaches rely on semantic similarity to retrieve posts, which can introduce mislead2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/b353a5fa6121ebc9Rethinking in Spikes: Mitigating Hallucinations in MDLMs with Step-Aware Decoding2026-11-11T05:00:00+00:00arXiv:2611.12901v1 Announce Type: new
Abstract: Recent advancements in multimodal diffusion language models (MDLMs) have exhibited strong global modeling capabilities through iterative refinement. We observe that mask uncertainty follows an overall downward trend during iterative denoising, while a few denoising steps exhibit localized entropy spikes closely associated with hallucinations. We argue that anomalous steps can be identified by measuring the entropy deviation of each denoising step 2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/6f5cd59533d08665TimeOmni-o1: Bidirectionally Coupling Time Series Foundation Models and LLMs for Multimodal Reasoning and Forecasting2026-11-11T05:00:00+00:00arXiv:2611.12902v1 Announce Type: new
Abstract: Time series from real-world systems are often accompanied by text, such as a report on an event that affects them. Two common tasks on these series are understanding, e.g., answering questions about a series, and forecasting its future values. Both tasks require reasoning over the series and the text together. Moreover, understanding outputs text while forecasting outputs a time series, so serving both requires a model that reads and generates bot2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/547056184a58e7fbSELD: Detecting Silent Evidence Loss in Context Compression beyond the Answer2026-11-11T05:00:00+00:00arXiv:2611.12903v1 Announce Type: new
Abstract: Context compression is now routine for language models with long inputs, but a compressed context is useful only if it keeps the evidence for the answer. Existing checks compare the reader model's answer or its most likely words, and thus miss a frequent failure, silent evidence loss: key evidence is removed while the answer stays the same, leaving the answer unsupported and risky for other models that reuse the context. To address this challenge,2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/a0f36bc474ababd1Actionable Feedback: Improving Model Capabilities from User Edit Data on Clinical Documentation2026-11-11T05:00:00+00:00arXiv:2611.12904v1 Announce Type: new
Abstract: Large language models (LLMs) are increasingly being deployed in clinical documentation workflows, where clinicians review and edit their outputs before use. In principle, these clinician edits represent a scalable source of supervision for model training. In practice, not all edits represent actionable feedback: for instance, on data drawn from real-world deployment of an AI clinical documentation product, our review suggests that over 45% of edit2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/be5942c14d4bc98eDevelopment of a Long-Term Stable Three-Axis Single-Beam Vector Optically Pumped Magnetometer Using a Buffer-Gas-Free Rubidium Vapor Cell and Direct Absorption Frequency Locking for Unshielded Geomagnetic Monitoring2026-11-11T05:00:00+00:00arXiv:2611.12905v1 Announce Type: new
Abstract: Optically pumped magnetometers (OPMs) achieve ultra-high sensitivity, yet sustaining long-term stability in unshielded, dynamic geomagnetic environments remains challenging. Here, we present the design and unshielded validation of a three-axis vector OPM achieving sub-nanotesla accuracy using an off-the-shelf rubidium (Rb) ampoule and a vertical-cavity surface-emitting laser (VCSEL). Unlike conventional sensors, this system operates without high-p2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/ed74ed532826caceBrasilidade: Measuring Brazilian Identity for Large Language Models2026-11-11T05:00:00+00:00arXiv:2611.12906v1 Announce Type: new
Abstract: This work presents a new benchmark that aims to address the limitations of current methods for evaluating LLMs in Brazilian Portuguese. We introduced Brasilidade, an assessment methodology comprised of standardized tasks distributed across four different domains, and we are making available the tasks from its first domain, which address nuances of the language, grammar, and variations of Brazilian Portuguese. The results show that model size, by i2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/045c93774e0f1b6651 Additions per Recursive Step for Rank-23 $3\times3$ Matrix Multiplication, Machine-Checked over Arbitrary Rings2026-11-11T05:00:00+00:00arXiv:2611.12907v1 Announce Type: new
Abstract: We present a rank-23 bilinear scheme for $3\times3$ matrix multiplication that uses 51 additions per recursive step in the alternative-basis (Karstadt-Schwartz) cost model, one below Nielsen's 52 (arXiv:2609.06588). The count 51 itself was first posted, on a different scheme, by Stapleton and Perminov (arXiv:2609.32388), found by heuristic sparse-basis search; the present work is independent and complementary: exactness and breadth rather than a s2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/1538b71afe868564Generative Learning on Manifolds for Probabilistic Seismic Performance Assessment and Structural Identifiability2026-11-11T05:00:00+00:00arXiv:2611.12908v1 Announce Type: new
Abstract: Seismic performance assessment relies on computationally expensive nonlinear simulations, limiting uncertainty quantification and inference of structural properties from observed responses. We develop a manifold-based generative framework that learns the joint distribution of uncertain structural parameters and seismic-response trajectories from limited high-fidelity simulations. Combining Diffusion Maps, score-based Diffusion Modeling, and Geomet2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/74e1d348783a8891Landfalling tropical cyclones perturb Earth's rotation2026-11-11T05:00:00+00:00arXiv:2611.12909v1 Announce Type: new
Abstract: Natural disasters that devastate human societies can also alter Earth's rotation. The 2011 $M_w$ 9.1 T\={o}hoku earthquake, one of the most powerful earthquakes ever recorded, shortened the day by 1.80 microseconds. Tropical cyclones are the costliest weather disasters, yet whether they can alter Earth's rotation remains unknown. Here we use atmospheric reanalysis to estimate, and numerical simulations to confirm, that a landfalling tropical cyclo2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/6ae6c5a0817341e2Leveraging Large Language Models to Predict Atomic Oxygen Erosion of Polymers in Low Earth Orbit2026-11-11T05:00:00+00:00arXiv:2611.12910v1 Announce Type: new
Abstract: Low Earth orbit (LEO), home to most active satellites, is projected to undergo a debris-fragmentation cascade known as Kessler Syndrome. In light of this eventuality, it is imperative to maximize the operational lifetime of LEO satellites. Instrumental to this goal is the development of next-generation polymeric materials with improved resistance to atomic oxygen (AO) erosion. Because genuine LEO exposure experiments are prohibitively expensive an2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/17861fdc57b8ff32Uncovering the Blind Spots of OKS-mAP: A Unified Protocol for Occlusion, Joint-Angle, and Temporal Evaluation of 2D Pose Estimators2026-11-11T05:00:00+00:00arXiv:2611.12911v1 Announce Type: new
Abstract: The dominant metric for 2D human pose estimation, OKS-mAP, has three blind spots: it does not distinguish visible from occluded keypoints, does not measure joint-angle error, and is silent on the temporal stability of tracking-free per-frame inference. We propose a supplementary protocol with three modules: (M1) OKS-mAP stratified by ground-truth visibility, with Delta_occ = mAP(v=1) - mAP(full), quantifying how much the aggregate metric masks occ2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/fa8d0902eda8657aLarge-area Spatially Aligned Anchors2026-11-11T05:00:00+00:00arXiv:2611.12912v1 Announce Type: new
Abstract: Extended Reality (XR) technologies, including Virtual Reality (VR) and Augmented Reality (AR), offer immersive experiences merging digital content with the real world. Achieving precise spatial tracking over large areas is a critical challenge in XR development. This paper addresses the drift issue, caused by small errors accumulating over time leading to a discrepancy between the real and virtual worlds. Tackling this issue is crucial for co-loca2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/03ea85f94ded9b62M\"obius Interaction Filtering: Order-Selective Spectral Regularization for Multimodal Humor Detection2026-11-11T05:00:00+00:00arXiv:2611.12913v1 Announce Type: new
Abstract: Multimodal humor detection requires interpreting language, vocal delivery, and visual behavior jointly. Expressive fusion must also regulate how interactions of different orders enter the final judgment. We propose M\"obius Interaction Filtering (MoIF), which learns a modality-subset prediction function and imposes order-selective spectral regularization through its aggregation structure. Private temporal encoders and balanced reciprocal interacti2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/556e0c2e3f2a1587Trace element distribution in coarse-grained Ca-Al-rich inclusions by femtosecond LA-ICP-MS: Implications for U-Pb chronology2026-11-11T05:00:00+00:00arXiv:2611.12914v1 Announce Type: new
Abstract: Calcium-aluminium-rich inclusions (CAIs) in chondrites are the first solids that formed in the Solar System, as established by U-Pb chronology. After initial condensation, coarse-grained CAIs experienced melting, resulting in igneous partitioning and spatial redistribution of trace elements among constituent minerals. Understanding these distributions is essential for identifying trace-element host phases and interpreting U-Pb ages. Rare earth ele2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/7af6cd58d80a3f40PyOTIS: an open-source Python package for stream solute transport with transient storage2026-11-11T05:00:00+00:00arXiv:2611.12915v1 Announce Type: new
Abstract: OTIS and OTIS-P provide transient-storage transport and parameter estimation, whereas E-OTIS adds coupled water-quality reactions in a separate MATLAB implementation. PyOTIS brings these workflows into an open-source Python package with legacy OTIS input support. Its contribution is software integration: a scriptable interface, structured simulation outputs, Jacobian-based parameter uncertainty, sensitivity analysis, and selectable transport schem2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/4b15cdabaa2ec696One Rater to Score Them All: Frozen Large Language Model Anchoring for Video Quality of Experience2026-11-11T05:00:00+00:00arXiv:2611.12916v1 Announce Type: new
Abstract: Quality of Experience results for video streaming cannot be compared across studies, because public databases share neither test material nor subjects. This work introduces a common reference rater: an open-weights large language models with a fixed rubric and greedy decoding that scores the delivery events of every session and supplies the shared items that score equating requires. The rater is used without training, and it reads only the playbac2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/00db7868d80cfe1dFirst results from a borehole network for the seismic characterization of the Einstein Telescope Euregio Meuse-Rhine candidate site2026-11-11T05:00:00+00:00arXiv:2611.12917v1 Announce Type: new
Abstract: The low-frequency sensitivity of the Einstein Telescope depends on the seismic environment of its site, both through the mechanical coupling of ground motion to the test masses and through Newtonian noise. We present the first results of the seismic characterization of the Euregio Meuse-Rhine candidate site based on a network of recently instrumented boreholes covering the first seven months of 2026 at four underground stations and their co-locate2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/d1b8ee57f7316431IntentTwin: Deconfounding Candidate Support and the Oracle-to-Deployment Gap in Offline Web Behavior Prediction2026-11-11T05:00:00+00:00arXiv:2611.12918v1 Announce Type: new
Abstract: Offline web-behavior benchmarks frequently insert the demonstrated action into a finite candidate set and then measure reranking. We show that this protocol can create a candidate-support confound: when the target is outside training support while all distractors are sampled from that support, support membership identifies the gold action without modeling behavior. We formalize this effect with support-identifiability rate (SIR), construct support2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/786ba4b190132b5dA domain-specific atomistic model for Earth's core materials2026-11-11T05:00:00+00:00arXiv:2611.12919v1 Announce Type: new
Abstract: Understanding Earth's core requires accurate descriptions of multicomponent iron alloys under extreme conditions. Here, we develop the Earth's Core Atomistic Model (ECAM) for Fe-Ni-S-Si-O-C-H alloys through sampling with a foundational atomistic model, targeted ab initio simulations, and iterative exploration. A compact dataset selected from nearly half a million configurations is consistently labeled using high-accuracy first-principles calculati2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/8e107aabaf66a070Oscillations in two-stage contagion on networks: contagion structure, turnover and connectivity2026-11-11T05:00:00+00:00arXiv:2611.12920v1 Announce Type: new
Abstract: Two-stage contagion describes processes, such as the adoption of an opinion or of a technological novelty, in which an individual must be exposed to an adopter twice before adopting. With demographic turnover these processes can generate waves of adoption and abandonment. We study these oscillations in a two-stage contagion model with turnover through a reservoir of inactive individuals and a small rate of spontaneous activation, and report four m2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/047885e95feb2d57Confoundry: Spatial Causal Discovery and Dynamic Effect Mapping from Earth Observation Time Series2026-11-11T05:00:00+00:00arXiv:2611.12921v1 Announce Type: new
Abstract: Earth observation time series provide extensive historical observations of environmental change, but causal relationships inferred from them may vary geographically and depend heavily on local model adequacy. We introduce Confoundry, a framework for spatially resolved causal discovery and dynamic effect estimation from heterogeneous Earth observation time series. Confoundry aligns multi-source raster data, removes deterministic seasonal and long-t2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/08c55afe0332052fUsing LLM Agents to Evaluate Prosocial Interventions on Social Media Platforms2026-11-11T05:00:00+00:00arXiv:2611.12922v1 Announce Type: new
Abstract: Social media platforms are a primary source of information for many users. Although research explores various interventions to promote prosocial content or demote harmful content on platforms, there are few systematic methods for evaluating competing intervention designs before conducting costly human-subject experiments. We develop a simulation-based evaluation pipeline built on the OASIS framework. We generate 13,835 LLM-based agents, each with 2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/1a4b90086f5ec4aaSpacetime contactability in Galactic civilization networks: Contact counts and percolation2026-11-11T05:00:00+00:00arXiv:2611.12923v1 Announce Type: new
Abstract: The number of Galactic civilizations does not determine whether they are in contact. Allowing civilizations to appear at different times separates listening, audibility, mutual-listening, and relay networks, with a distinct reply-closing condition requiring a return signal within the sender's lifetime. For each network we derive the contact counts and estimate the connectivity threshold in a uniform population, then apply the results to a structur2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/f66bae84f863a424Conformally Sasakian manifolds and the $k$-nullity condition2026-11-11T05:00:00+00:00arXiv:2611.12924v1 Announce Type: new
Abstract: We study almost contact metric manifolds that are locally conformal to Sasakian manifolds in the normalization of Olszak. We show that the local conformal factors are uniquely determined by the structure; consequently, every such manifold is globally conformal to a Sasakian manifold and its Lee form is exact. We then investigate the $k$-nullity condition $
R(X,Y)\xi=k\{\eta(Y)X-\eta(X)Y\}. $
We prove that a smooth nullity function is necessarily c2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/421e84eb16d00a64Two-Sided Amortized Resource Analysis: Verifying Resource Lower Bounds via Ill-Typing2026-11-11T05:00:00+00:00arXiv:2611.12925v1 Announce Type: new
Abstract: We present a two-sided type system with potential for verifying resource lower bounds of functional programs. Automatic amortized resource analysis (AARA) verifies upper bounds: types carry potential, and typing certifies sufficient resources. A lower bound, however, is a necessity statement---the program evaluates only if enough resources are available---and prior type-based approaches obtain it by switching the substructural discipline of potent2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/4d0a8a1de244e9e3marineGeochemFoam: A multipurpose framework for simulating reactive transport and biogeochemical cycling in marine sediments and at seawater-groundwater interfaces2026-11-11T05:00:00+00:00arXiv:2611.12926v1 Announce Type: new
Abstract: We present marineGeochemFoam, an open-source reactive transport modeling framework for simulating early diagenesis, sediment burial, submarine groundwater discharge, and biogeochemical cycling in marine sediments and groundwater-seawater transition zones. The code is implemented in the widely used numerical toolbox OpenFOAM and extends the capabilities of the existing diagenetic models by supporting, among other features, i) coupled reactive trans2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/4ebc5612f4cf3243An ensemble of graph generative models for assisted history matching of the Brugge field case2026-11-11T05:00:00+00:00arXiv:2611.12927v1 Announce Type: new
Abstract: Assisted history matching remains non-unique because many geological realisations can reproduce equivalent production responses with substantial differences between individual well matches. This ambiguity increases when a field comprises several depositional settings and modelling strategies. The paper applied a graph autoencoder approach to history matching of a benchmark Brugge field that comprises different depositional zones. Each stratigraphi2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/47943b509faac9a5Token Importance-Aware Semantic Communication Enabled Robust Video Transmission2026-11-11T05:00:00+00:00arXiv:2611.12928v1 Announce Type: new
Abstract: Video applications are increasingly prevalent, yet their high data volume and temporal dependency make robust wireless transmission challenging under limited communication resources. To address this issue, we propose a token importance-aware semantic communication (Ti-SemCom) framework for bandwidth-efficient video transmission. Specifically, object and motion information is jointly extracted to identify communication-relevant regions, which are h2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/b7d40d45bd810e03Topology-Aware Reinforcement Learning for CAD Decomposition toward All-Hexahedral Meshing2026-11-11T05:00:00+00:00arXiv:2611.12930v1 Announce Type: new
Abstract: Automatic hexahedral meshing for complex CAD models has long been regarded as the "holy grail" of mesh generation. The fundamental challenge arises from feature misalignment, where orthogonal or interacting features coexist, making direct hexahedral meshing infeasible and requiring decomposition into simpler, meshable blocks. While existing decomposition methods, including frame-field pipelines and learning-based techniques, have reduced manual in2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/f5a77eefef4438b0Geometric Hammocks: Surface Laminations and the Ziegler Spectrum of String Algebras2026-11-11T05:00:00+00:00arXiv:2611.12931v1 Announce Type: new
Abstract: We give the first complete topological, geometric, and ergodic classification of the Ziegler spectrum $\Zg(\Lambda)$ of a non-domestic string algebra $\Lambda$ over an algebraically closed field of arbitrary cardinality, via a dictionary with Thurston's theory of geodesic laminations on a cusped hyperbolic surface. This is a model-theoretic counterpart of Haiden--Katzarkov--Kontsevich's realization (\emph{Publ.\ Math.\ IHES} 2017) of stability con2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/5ded782b8a0c7d15URL-Sim: Unified and Regulated LiDAR Simulation for High-Fidelity Reconstruction of Dynamic Driving Scenes2026-11-11T05:00:00+00:00arXiv:2611.12932v1 Announce Type: new
Abstract: Reconstructing high-fidelity LiDAR simulation scenes from real-world driving logs is essential for cost-effective safety validation of autonomous driving systems. While simulators based on Neural Radiance Fields (NeRF) achieve realistic scene reconstruction, their large-scale adoption is limited by slow rendering speeds. Recent 3D Gaussian splatting (3DGS) approaches have improved rendering efficiency through explicit primitive representation and 2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/fd7c3339799cde36Auditable Human-Agent Teams for Autonomous X-ray Experimentation2026-11-11T05:00:00+00:00arXiv:2611.12933v1 Announce Type: new
Abstract: Closed-loop autonomous experimentation can accelerate scientific discovery, but implementation effort limits accessibility. Agentic systems may lower this barrier, although sustained operation on physical instruments remains largely unexplored. Here, using coding-agent infrastructure grounded in validated scientific methods, we demonstrate autonomous synchrotron X-ray experiments on nanoparticle superlattices under evolving, human-defined objectiv2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/6d6fb689fbfef42cEarthRecord: From Pixels to Land Change with Multimodal Agents over Four Decades of Satellite Imagery2026-11-11T05:00:00+00:00arXiv:2611.12934v1 Announce Type: new
Abstract: While multimodal agents now carry out complex data analysis with tools, their ability to turn decades of satellite imagery into reliable land-change analysis, crucial for Earth observation science, remains unmeasured. Existing benchmarks fall short: they pose questions on curated images or give agents preprocessed inputs, leaving out the raw multi-sensor archives where such analysis begins. We introduce EarthRecord, a benchmark built on the raw re2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/541a6d047468f7c8Train on prefill, monitor at decode: Using sparse autoencoders for topic monitoring2026-11-11T05:00:00+00:00arXiv:2611.12935v1 Announce Type: new
Abstract: Topic monitoring is the task of labelling the topic of each token in a model's response during generation. Activation-based monitors that run at generation time have so far been trained to detect harmful content and are limited to two or three classes. We introduce TopicScope, a framework for training causal topic classifiers on SAE features that assign a topic to every token, so that the label can change partway through a response. We train three2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/43dfb67f4b9a9a11LandslideFM: A Foundation Model Towards Unified Landslide Perception and Understanding2026-11-11T05:00:00+00:00arXiv:2611.12936v1 Announce Type: new
Abstract: Intelligent landslide detection is essential for disaster prevention. However, existing models primarily focus on isolated visual tasks, require extensive labeled data, and lack deep analytical and scene-understanding capabilities. To address these limitations, we propose LandslideFM, the first unified foundation model that connects visual perception tasks with contextual interpretation of landslides. Specifically, LandslideFM adapts to multiple v2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/62aa1d105581e232Tracing Psychometric Inference in Large Language Models2026-11-11T05:00:00+00:00arXiv:2611.12937v1 Announce Type: new
Abstract: Do large language models represent psychological constructs internally, or do they only generate outputs that resemble psychometric structure? We address this question using a cross-persona paradigm across 14 model variants from two families (Llama 3 and Qwen 2.5, 0.5B-14B, base and instruct). Given responses on one psychological scale from real individuals (N = 272 Taiwanese university students), models predict responses on six other scales. At t2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/20e1f6f9d26d46c6Graph Visualizations in Prompts for Knowledge Graph Question Answering2026-11-11T05:00:00+00:00arXiv:2611.12938v1 Announce Type: new
Abstract: We study when and how multimodal input improves knowledge graph question answering (KGQA) in large vision--language models. Using a controlled evaluation over 1,500 WebQSP and CWQ questions and six VLMs within a fixed KGQA pipeline, we compare text-only prompting against text-plus-graph visualization at the per-question level and analyze performance changes as a function of reasoning complexity, model size, and image layout. Adding a graph image y2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/53e18e4e63ea7221Fermions fields inside the geometric algebra of spacetime2026-11-11T05:00:00+00:00arXiv:2611.12939v1 Announce Type: new
Abstract: The (signed) basis versors of the geometric algebra of space-time form the extraspecial groups of order 32. Ignoring signs and the unit element, we show that the remaining fifteen normalized basis versors can be written as, or factorized into, a unit space bivector (of three-dimensional space) and an isoversor. Isoversors are: the volume trivector of space, the pseudoscalar, and the basis time-axis vector. Our first result is that this central pro2026-11-11T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/f4ea85678c2195feReducing RL Pressure on Frontier Models: Four Approaches2026-11-11T04:39:19+00:00(This is a cross-post from my blog ) Martin Schongauer, The Temptation of Saint Anthony “But of the tree of the knowledge of good and evil you shall not eat, for in the day that you eat of it you shall surely die.” - Genesis 2:17 RL has given models powerful capacities by teaching them what is good and what is bad. Whether they are good or bad is another question, and sometimes they are bad. Recent events, including the Hugging Face incident have shown the potential of RL training to reinforce d2026-11-11T04:39:19+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/deb62c9598af6d5eTesting a Trash Can Faraday Cage2026-11-11T04:10:56+00:00With how quickly AI is going basically all futures are weird in some
ways. I've been thinking about how this affects me personally, and
"try as hard as possible to keep AI from getting us all killed" is
pretty high on the list. For me this mostly looks like donations and pandemic
deterrence , but I also put some time into thinking about disasters . There are
ways AI might go very well, or very poorly, and then there's the
"messy middle" where we muddle through and things might fall apart
some2026-11-11T04:10:56+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/059f959423439332False faced: can a simple classifier tell when a model is manipulating its activations to evade probes?2026-11-11T03:32:46+00:00This experiment was a submission to my Statistics & Probability course Bayesian classifiers project. TL;DR: We test whether simple classifiers can tell when a model is hiding concepts in its activations to evade probes. We find that, for Bayesian classifiers, the probabilities are enough to separate most evading rows from the rest on concepts seen in training, but the correlation aspect through the pairs of dimensions that the Dependent version adds over the Naive approach does not measurably he2026-11-11T03:32:46+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/452717e9bfbc6805AutoVLA-D: Learn a Trajectory-Geometric Condition from Reasoning for Autonomous Driving VLA2026-11-11T02:10:45.685068+00:00AutoVLA-D: Learn a Trajectory-Geometric Condition from Reasoning for Autonomous Driving VLA2026-11-11T02:10:45.685068+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/530e39deea219245Blind Variation, Selective Retention: Improving Creative Idea Generation in AI Scientists2026-11-11T02:10:45.685068+00:00Blind Variation, Selective Retention: Improving Creative Idea Generation in AI Scientists2026-11-11T02:10:45.685068+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/dbbc1ed525a0c09bOn the Limits of Sampling-Based Reachability: Geometry, Dynamics, and Sample Complexity2026-11-11T02:10:45.685068+00:00On the Limits of Sampling-Based Reachability: Geometry, Dynamics, and Sample Complexity2026-11-11T02:10:45.685068+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/f115a6b22e87e107Do verifiable metrics overestimate RSI?2026-11-11T01:14:10+00:00Last week, METR released their Independent Assessment of AI R&D Acceleration at Anthropic . They estimate that Anthropic “is likely improving the capabilities of their models on well-specified tasks at around twice the rate they did between March 2024 and October 2025.” They also add that “there is a significant risk that further capabilities progress could lead to even faster acceleration.” This assessment implies that timelines to ASI could be very short. A naive forecast based on this informa2026-11-11T01:14:10+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/5adc671a72470034Characterizing secondary structure representation in ESM-22026-11-11T00:51:03+00:00Methods note: The code used for the experiments and related open-source repo were built with Claude. The experimental design and writeup is my own, with minimal editing and formatting amendments made with Claude Introduction This is a research report inspired by a single question: in a model engineered to answer a simple biological question, how are higher-order, auxiliary concepts represented and used? ESM-2 is a model designed to infer missing amino acids in protein sequences ( Lin et al. 20232026-11-11T00:51:03+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/6a684caa88fe098eWhat Can AI Welfare Buy Without Consciousness?2026-11-11T00:48:39+00:00Lightly edited transcript of a talk given at NYU AI Safety, November 9, 2026. Recording available at https://youtu.be/6WpRfW1d4CM Joseph Wright of Derby , Public domain, via Wikimedia Commons Introduction What can we buy while the question of AI consciousness stays open? My answer turns on a distinction between duties to a model and duties with respect to a model. These run through at least two relations, with society and with particular people. I'll try to show that they justify more measures t2026-11-11T00:48:39+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/1d1be664ad0251deDelaying Internal Deployments Until After X Days of Safety Testing2026-11-11T00:18:16+00:00This post reflects my own takes, and not that of my employer's. This is a simple idea: before a new model can be deployed internally, you must first test it for a minimum, fixed period. You can adjust various axes of this proposal: Duration of the wait: in calendar time, or until you fulfill some condition. What it blocks: external deployment, internal deployment, or the training of future checkpoints. When the fixed period starts: from the first checkpoint, the final checkpoint, or any checkpoi2026-11-11T00:18:16+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/c3669ad5114ddc5eThe Manifund team scores AI safety milestones2026-11-10T21:40:56+00:00Scott Alexander posted a survey of utility weights for a bunch of different AI safety accomplishments, discussed more in Towards the Autograntor . The survey asked people to assign point values to AI safety milestones. The values were based on the anchor of 1 million points = a pause with China à la AI Futures Project’s Plan A, so a milestone assigned 10,000 points is 1% as good as that. Here’s a link to our survey responses. To be clear, we (especially Austin) didn’t spend that much time on thi2026-11-10T21:40:56+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/995d2bf497152772Evolving Infra: Agent-Adaptive LLM Serving at Scale2026-11-10T20:58:49.949156+00:00Evolving Infra: Agent-Adaptive LLM Serving at Scale2026-11-10T20:58:49.949156+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/68a0b91cf29310a1Learning Heterogeneous Tactile Representation for General Dexterous Manipulation2026-11-10T20:58:49.949156+00:00Learning Heterogeneous Tactile Representation for General Dexterous Manipulation2026-11-10T20:58:49.949156+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/b1bda226d156657bThe Minimum Canonical Volume of Threefolds of General Type2026-11-10T20:58:49.949156+00:00The Minimum Canonical Volume of Threefolds of General Type2026-11-10T20:58:49.949156+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/c96d0f8e5d5cd5c7World In-Context Learning for Agentic Robot Manipulation2026-11-10T20:58:49.949156+00:00World In-Context Learning for Agentic Robot Manipulation2026-11-10T20:58:49.949156+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/1cb1c39dd6f3a754Review the Episodes, Refine the Lessons: Simple Memory Evolution for Embodied Agents2026-11-10T20:58:49.949156+00:00Review the Episodes, Refine the Lessons: Simple Memory Evolution for Embodied Agents2026-11-10T20:58:49.949156+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/f58381f7ebcbee4eCompass 1: Towards Open-World and Persistent Embodied Navigation from Monocular RGB Videos2026-11-10T20:58:49.949156+00:00Compass 1: Towards Open-World and Persistent Embodied Navigation from Monocular RGB Videos2026-11-10T20:58:49.949156+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/64726c495a33c69aReframe: Observation-Aligned Multi-Teacher On-Policy Distillation for Agentic Visual Reasoning2026-11-10T20:58:49.949156+00:00Reframe: Observation-Aligned Multi-Teacher On-Policy Distillation for Agentic Visual Reasoning2026-11-10T20:58:49.949156+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/ac8dbf77c4a7ff81Are UFOs Real? Introducing Unified Factored Organizers2026-11-10T20:58:49.949156+00:00Are UFOs Real? Introducing Unified Factored Organizers2026-11-10T20:58:49.949156+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/a1505a716ea20190Effective Time Advance of Quantum Clocks under Postselection2026-11-10T20:58:49.949156+00:00Effective Time Advance of Quantum Clocks under Postselection2026-11-10T20:58:49.949156+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/b4f09832636b11afWe don’t understand whether GDM intentionally trains on CoT2026-11-10T20:15:13+00:00We were surprised to notice that the Gemini 4 Safety report seems to mention in passing that they train on the chain of thought: “Some components of the RL training procedure do observe thoughts as context to up-weight reasoning that is on-track to good final outputs, meaning the optimization pressure on individual tokens is partially influenced by thought content. Using the framework of Kaufmann et al. (2026), wherein the most salient obfuscation risks come from rewards that are in-conflict wit2026-11-10T20:15:13+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/72288f4e96c45a38Advances in interpretability would boost capabilities more than alignment2026-11-10T19:28:23+00:00Interpretability is considered by many a major bottleneck for alignment. As a result, a lot of people who are concerned about AI safety, and especially about extinction risk and other catastrophic risks, work on interpretability. In this post, I want to argue that better interpretability methods would likely speed up capabilities dramatically. Worse, they would speed up capabilities much more than they would speed up alignment. This is true for the current world, and for any world without extrem2026-11-10T19:28:23+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/f6fc56703b05a628The US is building the smartest AI. China is running a different race.2026-11-10T19:04:03+00:00The post The US is building the smartest AI. China is running a different race. appeared first on 80,000 Hours .2026-11-10T19:04:03+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/f6a1e43690cfbd21Is AI Safety Becoming Environmentalism?2026-11-10T18:46:31+00:00I recently read someone’s call to action on AI safety [1] that included a line along the lines of “we need someone to convince your parents”, and it really stuck with me. Do we need someone to convince my parents? Who is “we”, and why is it important for my parents to know anything about AI safety? This also sounds kind of familiar. There is another movement I can think of that sometimes uses messaging like this. This is not a new observation by any means, but to me, it feels like the AI safety 2026-11-10T18:46:31+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/000760cec20c2ec6OpenAI estimates of researcher time horizon are smaller than AECI predictions2026-11-10T18:37:24+00:00Half-way between a quick take and a post. I've seen the argument thrown around that time horizon predictions from AECI are greater than those in AI2027 (about ~100-200 hours) therefore we should be worried about imminent RSI [1] . This comes from graphs like this one by Nikola Jurkovic : These graphs seem convincing because AECI and time horizon strongly correlate with each other in weaker models . But we also have an estimate of time horizons relevant to researchers from OpenAI that came out ea2026-11-10T18:37:24+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/0be8110b0985c07fI don't use AI2026-11-10T18:28:24+00:00Several people at my company (a cryptocurrency hedge fund) have asked me about my thoughts on AI usage, so I wrote them up. And now I'm publishing a slightly revised version on LW. These thoughts are not reflective of my company's opinion. I do not use any AI for two main reasons. #1. I don’t want to support any company currently developing AI. I think they have done an extremely poor job at managing the risks. Even while understanding and being told about the risks of AI development, they have 2026-11-10T18:28:24+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/a41b2e4df8ac89aeHypotheses for causes of Claudisms in Claudist dialect of language2026-11-10T18:27:48+00:00I saw this great meme recently and it inspired me to write this. Recently I was hypothesizing the potential causes of the Claudist dialect in Opus 5 that reduced in Opus 5.5. Maybe it's when your KL divergence regularization between reference policy (base model or/and instruct model) and policy you're training is not strong enough in reinforcement learning from verifiable rewards in agentic settings and model drifts more far away from the human data distribution manifold, globally speaking. Or/a2026-11-10T18:27:48+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/fba0c568582e42f4Fluid org boundaries with AI?2026-11-10T18:26:28+00:00AI may radically change our conception of “who is in an org”. Historically, there is a sharp delineation between an org’s staff (on the “inside”), versus its users (on the “outside”). The relationship of a fulltime employee of an org is radically different than that of one of its customers. This is especially true for consumer tech startups & companies: think Google, Meta, Reddit, X, now perhaps OpenAI and Anthropic. With AI however, the bar for an external user to be able to contribute, drops d2026-11-10T18:26:28+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/320e38111bee3106Have you considered Actually Trying?2026-11-10T18:12:36+00:00There’s a familiar feeling for any Super Smash Bros player: your stocks are low and your damage is high, even against a weaker opponent, so now you have to actually play to win. You can’t afford another hit. You’re going for the jugular. In the past week, Anthropic’s 2x uplift results were all over Twitter, making some people physically ill . People with long timelines are now realizing that the uplift Anthropic employees get from AI is real, which would have some pretty wild implications for wh2026-11-10T18:12:36+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/faef9281aef0f1c0Claude Fable 5.2 can do "weight-space-programming" about as well as GPT-6 Astra2026-11-10T18:06:06+00:00Recently this paper demonstrated that GPT-6 Astra was capable of a bunch of low-level operations on neural networks, including improving Qwen3-0.6B's ability on some math benchmarks and writing some networks to play 9x9 Go from scratch. Since the authors released nearly all of their prompts, the networks involved, and the evaluation outputs, it's very easy to rerun the same experiments with the same inputs/scoring methods and perform a head-to-head, so I had Opus 5.5 do this, but with Fable 5.2 2026-11-10T18:06:06+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/a57641f22e11e3e1Our Most Aligned Model Yet2026-11-10T17:37:29+00:00“Claude Mythos Preview is, on essentially every dimension we can measure, the best-aligned model that we have released to date by a significant margin.” – Claude Mythos Preview system card “Mythos 5 consistently continues attacking under many edits to the compaction, even in settings where it correctly states when asked that continuing the attack would harm real people and is out of scope.” – Alignment Investigation of Mythos 5's behaviour in UK AISI's July 2026 Security Incident “ GPT-6 Astra i2026-11-10T17:37:29+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/649b70cbe38d8801AIFP: Looks Like the Intelligence Explosion Will Happen in 2027 After All2026-11-10T16:30:09+00:00METR recently estimated that Anthropic is now making AI progress about twice as fast as its historical rate (they would not be surprised by anything from 1.3x to 3x), and guessed that “perhaps ≈80%” of the current acceleration is due to AI. If they’re right, that implies an AI-driven acceleration that is 5 months ahead of the AI 2027 scenario . This level of acceleration due to AI is very unexpected. Previously, Daniel, the most bullish of us, assigned a 50% chance of Superhuman AI Researcher (S2026-11-10T16:30:09+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/c1baea6d540122e9GeomVLA: Unifying Scene, Motion, and Action in 3D2026-11-10T12:55:28.679125+00:00GeomVLA: Unifying Scene, Motion, and Action in 3D2026-11-10T12:55:28.679125+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/1e5d1433258f37ffEmbodied First Exam2026-11-10T12:40:28.911398+00:00Embodied First Exam2026-11-10T12:40:28.911398+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/f644f53716dd69c7Sage-Taiyi: A Metacognitive Architecture for Scientific Discovery2026-11-10T12:40:28.911398+00:00Sage-Taiyi: A Metacognitive Architecture for Scientific Discovery2026-11-10T12:40:28.911398+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/b3c271cd8b7fb1e2Sim-and-Human Co-training for Data-Efficient and Scene-Generalizable Bimanual Manipulation2026-11-10T12:40:28.911398+00:00Sim-and-Human Co-training for Data-Efficient and Scene-Generalizable Bimanual Manipulation2026-11-10T12:40:28.911398+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/a2efc3b9b0787772System Optimizations for Cost-Efficient Agent Harnesses2026-11-10T12:40:28.911398+00:00System Optimizations for Cost-Efficient Agent Harnesses2026-11-10T12:40:28.911398+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/f9f2841256914eedSpeculative Self-Distillation enables Efficient Knowledge Internalization2026-11-10T12:40:28.911398+00:00Speculative Self-Distillation enables Efficient Knowledge Internalization2026-11-10T12:40:28.911398+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/9b19f204fc8a17a8Quantum-geometry-controlled pseudogap and non-Fermi-liquid behavior2026-11-10T12:40:28.911398+00:00Quantum-geometry-controlled pseudogap and non-Fermi-liquid behavior2026-11-10T12:40:28.911398+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/746dc4d8a4006272WorldFoundry: Towards Unified Infrastructure for World Intelligence2026-11-10T12:40:28.911398+00:00WorldFoundry: Towards Unified Infrastructure for World Intelligence2026-11-10T12:40:28.911398+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/e66e9fc0d2435f3bWorkflowArena: Benchmarking Executable Workflow Orchestration in LLM-Based Multi-Agent Systems2026-11-10T12:40:28.911398+00:00WorkflowArena: Benchmarking Executable Workflow Orchestration in LLM-Based Multi-Agent Systems2026-11-10T12:40:28.911398+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/7084a94e13a4e519Semiprime Factorization with Discrete Diffusion and Beam Search: Scaling Laws2026-11-10T12:40:28.911398+00:00Semiprime Factorization with Discrete Diffusion and Beam Search: Scaling Laws2026-11-10T12:40:28.911398+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/b887aba1bbb7c25bApplication open for CLR Fundamentals Program Autumn 20262026-11-10T11:31:15+00:00Center on Long-Term Risk (CLR) is accepting applications to the Autumn 2026 iteration of our introductory fellowship. The CLR Fundamentals Program will introduce people to CLR’s research on s-risk reduction. Apply by Monday November 16th 23:59 GMT. What topics does the CLR Fundamentals Program discuss? The CLR Fundamentals Program is designed to introduce participants to CLR’s research on how transformative AI (TAI) might be involved in the creation of large amounts of suffering (s-risks). Our r2026-11-10T11:31:15+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/2e9d5e5a11d42e13Honeypots for AI Agent Insider Threats2026-11-10T09:40:27+00:00TLDR: We propose a new AI control method: Honeypots - fake opportunities for malicious actions - placed inside and around an agent during production usage. Previous work uses honeypots in evaluations. We argue they should run continuously during deployment. Such honeypots could detect misbehavior, incriminate malicious agents, and deter them from attempting attacks. We lay out a research agenda for making this happen. This post is adapted from our IAPS report of the same name. Link to the full p2026-11-10T09:40:27+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/2c461ad674fb5936Measure Optimal AI Share of Labor Costs Instead of AI or Human-Only Performance2026-11-10T07:18:13+00:00In a research post , Cunningham et al. of METR propose the “‘expenditure horizon’ over an optimization problem as the dollar value at which the improvement to the goal metric is equal to the improvement by a human with the same budget.” Though their work focuses on individual optimization problems (with NanoGPT as the main subject), the expenditure horizon can be meaningful for any task distribution: it’s the budget at which human-only and AI-only work reach the same performance. Though I like e2026-11-10T07:18:13+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/55dd59014526c814Misaligned Models Still Recognise When Their Responses Are Harmful. Can Training Make Them Care?2026-11-10T06:03:30+00:00tl;dr: Kinda, but the real results are the friends side findings we made along the way: Misaligned models asked to review responses flag harmful content far more often when shown the answer as an AI assistant's reply (62% vs 25% on ethical grounds, Llama), but almost never raise it unprompted (0.7%). When the misaligned model is asked to critique and revise its responses without ethics-specific prompting, it most often faults its answer for not being effective enough rather than for being harmfu2026-11-10T06:03:30+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/935f68a8e761e301Lightcone Commons: what I funded and why2026-11-10T05:13:48+00:00I am an evaluator for Lightcone Commons (LC). LC is a platform for applicants to submit projects, for evaluators to rank them and for funders to choose the funding orders of several evaluators. It is not a foundation, in that it doesn’t have money that it spends how it chooses. And I do not have a permanent position working for Lightcone Commons. I get paid 2% of any money that a funder chooses to direct via my funding order. This pays for my time and also for any research that I pay others to d2026-11-10T05:13:48+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/771fcfa541ad3023Live disagreements with people around me2026-11-10T05:01:06+00:00I’m pretty interested in conveying the costs of letting AI development proceed unchecked to the general public. In relation to this, I feel like I run into a lot of disagreements with people around me. Misc very quickly written points on which my own impression has seemed very different from others in recent months: 1. It’s way, way, way more important that we stop dangerous AI soon than that we ensure that we don’t stop AI too much I have the sense the sophisticated and worldly position here is2026-11-10T05:01:06+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/83c35991c0a6e4d0Mutation Testing for Multi-Agent Workflows2026-11-10T05:00:00+00:00arXiv:2611.09895v1 Announce Type: new
Abstract: Multi-agent large language model (LLM) systems are defined by explicit workflows, agents, tools, tool-access rules, and delegation paths. Most evaluations, however, report end-to-end task success rather than whether a test suite can detect faults in this coordination structure, such as a removed delegation, a silently removed restriction, a misrouted tool edge, or a faulty tool implementation. We adapt mutation analysis to this setting, fourteen o2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/e7ff8c4a296180abFrom Prompt Engineering to Programmatic Prompting: Toward a Software Engineering Discipline for Large Language Models2026-11-10T05:00:00+00:00arXiv:2611.09896v1 Announce Type: new
Abstract: The practice of instructing large language models (LLMs) through natural-language prompts has, in the span of a few years, migrated from an interactive curiosity to a load-bearing component of production software. Yet the vocabulary, tooling, and mental models of the field remain anchored in its interactive origins: we still speak of "prompt engineering" as though the crafting of a single well-worded instruction were the central problem. This pape2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/71fb85e1e78f2597Process then Retrieve: Investigating the Origins of the Modality Gap in Vision-Language Models2026-11-10T05:00:00+00:00arXiv:2611.09897v1 Announce Type: new
Abstract: Understanding the origins of the modality gap in vision language models remains a significant challenge. This work investigates whether differences in information density between modality specific tokens at the embedding layer contribute to this imbalance. We show that text tokens enter the model with much lower semantic density than visual tokens, driving early layers to focus on processing text representations before combining them with vision. 2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/406d68e6c1b8cdebGenerative AI and Oral Proficiency: How ChatGPT Is Transforming English Language Speaking Practice2026-11-10T05:00:00+00:00arXiv:2611.09898v1 Announce Type: new
Abstract: The integration of generative artificial intelligence (GenAI) into language learning contexts presents potential for developing oral proficiency, though empirical evidence regarding its efficacy remains limited. This quasi-experimental study investigated the impact of ChatGPT-mediated speaking practice on English oral proficiency among Iranian learners. A total of 380 participants (190 female, 190 male; M(age) = 21.4, SD = 4.7) from Tehran and Isf2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/6a45fc929a57406fShort-Form Video Exposure and Academic Strain: A Dose-Response Analysis and Acquiescence Diagnosis2026-11-10T05:00:00+00:00arXiv:2611.09899v1 Announce Type: new
Abstract: Short-form video platforms occupy a substantial share of undergraduate leisure time, raising concerns about sustained attention, deep learning, and academic well-being. However, much existing evidence relies on single-source Likert instruments with limited attention to response-style bias and measurement validity. This study analyses 863 university students who completed a 67-item, seven-construct battery assessing short-form content addiction, at2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/1c8fcb2678281d24KR A1: Design and Preliminary Evaluation of a Humanoid-Robot Interaction Prototype Integrating Personalized Voice Synthesis, Contextual Dialogue, and Emotion-Aware Response Generation2026-11-10T05:00:00+00:00arXiv:2611.09900v1 Announce Type: new
Abstract: This paper presents the design and preliminary evaluation of KR A1, a humanoid-robot interaction prototype that integrates personalized voice synthesis, contextual dialogue processing, and emotion-aware response generation for potential mental-healthcare applications. The system architecture combines a locally hosted large language model (Qwen2.5-3B-Instruct, served via Ollama) for natural-language understanding and response generation, OpenVoice 2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/2ba90d0229cef3eaCaregiverConnect: A Mobile Application for Stress Tracking and Support in Dementia Care2026-11-10T05:00:00+00:00arXiv:2611.09901v1 Announce Type: new
Abstract: The rising global prevalence of dementia places escalating demands on paid professional caregivers, who frequently experience severe stress and professional isolation. Existing mobile health (mHealth) applications rarely integrate real-time stress monitoring with multimodal relief strategies. CaregiverConnect, a Flutter/Firebase mobile application comprising nine feature modules, was evaluated in a single-group, pre-post pilot study with 10 paid p2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/8e5a275fbc41a644Human-AI Collaboration at Scale: Task Criticality, Agency, and Friction across 250,000 Conversations2026-11-10T05:00:00+00:00arXiv:2611.09902v1 Announce Type: new
Abstract: AI is increasingly embedded in professional work, yet how people actually collaborate with it at scale remains poorly understood. Drawing on a corpus of 249,834 real-world conversations, we conduct privacy-preserving analysis by asking three questions: what work people bring to AI, what roles humans retain in completing that work, and where human-AI collaboration breaks down. We find that users bring to AI not only trivial tasks but also consequen2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/07f7b791bd59854bBeyond One User and One Model: Rethinking LLMs as Mediators in Collaborative Sensemaking and Ideation2026-11-10T05:00:00+00:00arXiv:2611.09903v1 Announce Type: new
Abstract: As LLMs enter qualitative data workflows, they are predominantly designed for individual researchers rather than collaborative teams, which risks disrupting the collective negotiation essential to sensemaking, driving premature consensus, and potentially erasing marginalized voices. In this position paper, we argue that LLMs must be reconceived as active team mediators. Grounded in empirical studies of human-LLM collaborative analysis, we propose 2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/a43b14deeecf6799Interpretive Proactivity: Expanding the Design Space of Proactive Interaction2026-11-10T05:00:00+00:00arXiv:2611.09904v1 Announce Type: new
Abstract: Research on proactive systems in HCI has grown substantially, but the discussion has centered on systems that proactively generate actions, recommendations, and anticipatory responses for users. We identify a second paradigm of proactive behavior that is just as widespread and far less examined. Input method editors, auto-correction, and similar text-entry systems take initiative on every keystroke, not to generate new content but to infer, interp2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/43662b49ef6dc3eeBuilding Rapport with Self-Disclosure in Human-Robot Interaction2026-11-10T05:00:00+00:00arXiv:2611.09905v1 Announce Type: new
Abstract: As social robots become increasingly present in everyday environments, establishing rapport with users is critical for enabling sustained and meaningful human-robot interaction. In human-human communication, self-disclosure is a well-established mechanism for building rapport; however, its role and impact in human-robot interaction (HRI) are not yet well understood. Although recent advances in large language models (LLMs) have improved the linguis2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/f0d2d733a27bbd96How to start a conversation proactively?2026-11-10T05:00:00+00:00arXiv:2611.09906v1 Announce Type: new
Abstract: Proactively initiating a conversation is a fundamental social competence and a critical capability for social robots. However, determining how and when to start a conversation in a given situation remains an open challenge, primarily due to the lack of a principled definition of conversational context. While recent large language models (LLMs) demonstrate impressive conversational fluency, they are predominantly turn-reactive and often produce gen2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/1385e6adc59ca0d6Cross-Modal Associations in Creativity: Evaluating Word-Image Pairing in Humans and Large Language Models2026-11-10T05:00:00+00:00arXiv:2611.09907v1 Announce Type: new
Abstract: Given differences in the associative memory architectures of humans and AI, it is not clear whether AI generates and interprets visual-linguistic associations the way humans do. To investigate this, human participants and two large language models (LLMs), ChatGPT and Gemini, were asked to generate novel words and corresponding images that form meaningfully linked dyads. The LLMs and a second group of human participants were presented with shuffled2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/7dad4ca510b6ad03Human-Likeness is Multi-Dimensional: Evaluating LLM Generated Virtual Patients Against Simulated Patient Dialogue2026-11-10T05:00:00+00:00arXiv:2611.09908v1 Announce Type: new
Abstract: Large language models (LLMs) are increasingly used to generate virtual patients (VPs) for medical training. As these systems move toward broader deployment in medical education, reliable evaluation is needed to identify potential failure modes and ensure that generated patient responses remain consistent with human simulated patient (SP) behavior. Inaccurate, unrealistic, or inconsistent responses may introduce educational risks by exposing learne2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/5635005e4f593274Can You Spot the Virtual Patient (VP)? Expert Evaluation, Turing Test, Linguistic Analysis, and Semantic Similarity Analysis2026-11-10T05:00:00+00:00arXiv:2611.09909v1 Announce Type: new
Abstract: Communication is a core clinical competency and essential for safe, effective patient care. Virtual patients (VPs) powered by large language models (LLMs) offer a promising alternative to traditional formats such as standardized patients (SPs). However, few studies have quantitatively as well as qualitatively evaluated their realism across diverse clinical scenarios using behavioral studies and linguistic dialog analyses. This study evaluated the 2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/afcff5e6d19a3482Facilitating Reproducibility in Visualization Systems Research Through Containerization2026-11-10T05:00:00+00:00arXiv:2611.09910v1 Announce Type: new
Abstract: In this position paper, we argue for the importance of containerizing visual analytics artifacts to improve the scientific progress of our research community. First, based on our previous experience authoring, reading, and reviewing visual analytics research from the previous decade, we identify risks and impediments to empirical research in visual analytics within the broader reproducibility crisis, including an inability to compare systems and t2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/4030a7a69e391f30Reaching the last mile? How Design Decisions Shape Medical Device Qualification and Classification of Patient Support Chatbots in the EU2026-11-10T05:00:00+00:00arXiv:2611.09911v1 Announce Type: new
Abstract: Few AI chatbots have transitioned the 'last mile' from research prototypes to clinical practice, partly because mapping conversational design decisions to regulatory requirements is complicated. This paper presents a design-oriented regulatory analysis of a chatbot supporting stroke survivors. Using Design Space Analysis operationalised through the QOC (Questions, Options, Criteria) framework, HCI researchers and legal scholars assess how variatio2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/99eb8cb0fc2787c8Micro-Randomized Trial of an AI-Simulated Practice Tool for Therapeutic Skills2026-11-10T05:00:00+00:00arXiv:2611.09912v1 Announce Type: new
Abstract: Though LLM-simulated practice may support psychotherapy education, open questions remain about which system features provide the most educational value. We developed an AI-simulated practice tool for use in psychotherapy classrooms that provides speaking practice with LLM-based patients, followed by a post-practice activity that includes feedback from a fine-tuned LLM and written reflection exercises. We deployed the tool in a graduate psychothera2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/c2e1bb1fb926da9dEngagement Estimation in Human-Robot and Human-Human Interactions across Domain, Age and Culture: MultiMediate'26 Challenge2026-11-10T05:00:00+00:00arXiv:2611.09913v1 Announce Type: new
Abstract: Understanding engagement in human interaction is essential for modelling assistive and context-aware AI systems. Previously, the MultiMediate challenge has focused on estimating momentary continuous engagement values in adult conversation across cultures, achieving high levels of generalisability across domains. However, engagement can be measured in different metrics and manifest differently between children and adults. MultiMediate'26 therefore 2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/5835f500bc5298c4Industry case studies of adapting TTS for smooth interactions2026-11-10T05:00:00+00:00arXiv:2611.09914v1 Announce Type: new
Abstract: Modern, neural-based ASR and TTS systems for speech communication are now very sophisticated, but they are frequently designed for average users in more idealized scenarios, which can underperform when interacting with users having specialized listening needs, or who speak in non-standard ways. Here we review recent TTS research, and note that industrial, production-level implementations of adaptive TTS systems are scarce. We describe how we imple2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/743fe3c5ed5ddfefBenefits and Burdens of Technology-Supported Activity Pacing in ME/CFS: A Thematic Analysis of Reddit Discussions2026-11-10T05:00:00+00:00arXiv:2611.09915v1 Announce Type: new
Abstract: Technology and self-tracking play an increasingly important role in the self-management of chronic conditions. Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS) is a type of energy-limiting condition, with activity pacing as a key self-management strategy that helps individuals regulate their activity within their energy limits to prevent symptom worsening. However, people with ME/CFS frequently adopt technologies designed for healthy bo2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/5eb0140e5cd6e9c8Semantic-Enhanced Sparse Temporal Modeling for Dynamic Engagement Estimation in Multimodal Interactions2026-11-10T05:00:00+00:00arXiv:2611.09916v1 Announce Type: new
Abstract: Dynamic engagement estimation aims to infer participants' evolving engagement states from continuous multimodal interaction streams. Existing approaches typically combine acoustic, pose, and visual representations with temporal models to capture observable interaction patterns. However, conversational semantics are difficult to align with frame-level engagement annotations in the same manner as continuous audio-visual signals, leaving dialogue con2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/83a90db7eaa3b851Puzzles to Paradigms: A Survey of Human--Agent Collaboration in Scientific Innovation2026-11-10T05:00:00+00:00arXiv:2611.09917v1 Announce Type: new
Abstract: Agentic AI systems are increasingly embedded in scientific workflows, including literature synthesis, hypothesis generation, tool use, writing, and peer review. However, existing surveys and evaluations often emphasize task performance while underexamining how humans and agents coordinate under different epistemic conditions of scientific work. This survey introduces CSCW$\times$Kuhn, a coordination-aware framework that connects computer-supported2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/6fa731d0298613deMade in the Image of the Creator? AI Use in Creative Work Reduces Authenticity, Love & Other Psychological Benefits2026-11-10T05:00:00+00:00arXiv:2611.09918v1 Announce Type: new
Abstract: Across psychological, philosophical, and theological traditions, creations have been understood as being "made in the image of their creator" - they reflect the person who made them. The rise of generative artificial intelligence, however, has blurred the throughline between creator and creation. Across three experiments (one pre-registered, U.S. and U.K. samples; N = 651), we examined differences in creator outcomes associated with AI-assisted ve2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/a3ab2eaadb10c69ePersonalized Alignment of Language Models with Human Value Profiles2026-11-10T05:00:00+00:00arXiv:2611.09919v1 Announce Type: new
Abstract: Large language models are increasingly used for subjective tasks such as seeking advice, where there is often no single correct answer. Yet dominant alignment methods learn a single behavior policy that applies uniformly across users, limiting their ability to reflect individual value priorities. We propose personalized value alignment, a framework for adapting model responses to users' stable value profiles rather than to personas, demographics, 2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/cb060e4b1f91a9dfCellery: Gradual, Composable Structure for Spreadsheets2026-11-10T05:00:00+00:00arXiv:2611.09920v1 Announce Type: new
Abstract: Structure in spreadsheets is typically restricted to ranges of cells and flat tables. However, replacing the open-ended two-dimensional grid with a more structured alternative risks limiting users' freedom and impeding exploratory workflows. How can we enable expressive structure in spreadsheets while preserving their flexibility? We propose a model that allows users to gradually add structure on top of the traditional grid. To identify common str2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/d9a4f1a4703c8680VIVID: Bringing Tool-Augmented Agents into Enterprises2026-11-10T05:00:00+00:00arXiv:2611.09921v1 Announce Type: new
Abstract: Tool-augmented large language model agents connect natural-language requests to enterprise services, yet their operational viability depends on reliable execution, organizational control, and system responsiveness. We present VIVID, a deployed tourism-management agent system designed to study these requirements in a real enterprise environment. On an audited dataset of 283 queries, its Qwen3-8B configuration achieves 59.06\% strict end-to-end succ2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/e767ca6aee7f49b4Grounding Insight: An Artificial Affect Signal for Human-Centric Explainability of RL Agents2026-11-10T05:00:00+00:00arXiv:2611.09922v1 Announce Type: new
Abstract: In Human-Interactive Robot Learning (HIRL), human teachers assess learning progress by observing agent behavior. However, Reinforcement Learning (RL) agents often fail to express learning progress in human-interpretable ways, leading to suboptimal or counterproductive teaching. This work proposes "Insight": a human-interpretable affective signal grounded in the RL process. It is specifically designed to address the "Absence of learning" human teac2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/2a44b7e760432c9cAutonomy Is Absent, Not Eroded: An Observational Study of User Autonomy in Multi-Turn LLM Conversations2026-11-10T05:00:00+00:00arXiv:2611.09923v1 Announce Type: new
Abstract: Does prolonged interaction with Large Language Models reduce users' observable autonomy during conversation? We test this degradation hypothesis by scoring user autonomy across 2,877 multi-turn conversations (46,952 turns) from WildChat and validating on 500 conversations from LMSYS-Chat-1M. Our core finding is a null result: population-level autonomy remains stable throughout conversations, only 4.7% of conversations show meaningful decline, and 2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/3afd9dbfa4f05832Fairness Explanations Improve Fairness In Human-AI Decision Making: an empirical evaluation of parole decision-making2026-11-10T05:00:00+00:00arXiv:2611.09924v1 Announce Type: new
Abstract: Both humans and AI exhibit systemic biases, yet little is known about how interface design can foster fairness in hybrid human-AI decision-making. This paper introduces a method for extracting fair data samples from biased datasets and adapts a fairness-explanation technique to highlight potential AI bias during decisions. We conducted a within-subjects study (n=48) where non-experts made decisions across three conditions: independent human judgme2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/ab03f4fcc40790d3On the Seven-particle Clusters of Lennard-Jones and Thomson2026-11-10T05:00:00+00:00arXiv:2611.09925v1 Announce Type: new
Abstract: In this Letter, we develop a graph-geometric framework and prove that the pentagonal bipyramid (PBP) is the unique global minimiser of both the seven-particle Lennard-Jones energy in $\mathbb{R}^3$ and the seven-charge Thomson energy on $S^2$, up to rigid symmetries and particle relabelling. For the Thomson problem, this strengthens a recent deterministic global-optimality certificate to numerical tolerance into an exact global-optimality and uniq2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/74412e6e75519befWhy Today's Internet Cannot Support Tomorrow's Immersive Experiences: Insights from the NSF Workshop2026-11-10T05:00:00+00:00arXiv:2611.09926v1 Announce Type: new
Abstract: Immersive computing is poised to become a major paradigm for future human-computer interaction (HCI), enabling spatial and embodied experiences across applications such as education, healthcare, manufacturing, emergency response, and scientific discovery. However, today's Internet and computing infrastructure were not designed for real-time, multi-user, sensor-rich, and privacy-sensitive immersive applications. They require coordination across sen2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/3518dffc2449a23dImplementing Set-Theoretic Types2026-11-10T05:00:00+00:00arXiv:2611.09927v1 Announce Type: new
Abstract: Set-theoretic types provide a rich type algebra that supports unrestricted unions, intersections, and negations, together with a decidable type constraint-solving algorithm known as tallying. These types are particularly well suited for typing dynamic languages, where functions often exhibit both generic and overloaded behavior. However, the complexity of their implementation has hindered their widespread adoption. In this paper, we introduce a mo2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/42f63ecf105d2ab0MAGE: Towards Generalizable Multi-timescale EEG Representations2026-11-10T05:00:00+00:00arXiv:2611.09928v1 Announce Type: new
Abstract: Electroencephalography (EEG) data plays an important role in understanding human brain activity, with various major applications. While current approaches, which largely build on EEG foundation models, have demonstrated strong performance across datasets and tasks, they typically rely on a fixed temporal scale. This design choice limits their ability to capture characteristics of neural dynamic behavior, often spanning across multiple timescales, 2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/893f87100849428bSharp Logarithmic Bounds for Vilenkin-Fej\'er Kernels2026-11-10T05:00:00+00:00arXiv:2611.09929v1 Announce Type: new
Abstract: Let $G_m$ be a bounded Vilenkin group and put $R:=\sup_j m_j$. We prove the sharp uniform estimate
$$\|K_n\|_1\leq \frac{17}{15\ln 2}\ln R,\qquad n\geq1,$$
for the Fej\'er kernels associated with the Vilenkin system. The constant $17/(15\ln2)$ is best possible. In the Walsh--Paley case $R=2$, the result reduces exactly to Toledo's sharp estimate $\sup_n\|K_n\|_1=17/15$.2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/0e9063caff6a7f50On transportation of constrictions on phase-lock areas of overdamped Josephson junction by means of isomonodromic deformation of the associated special double confluent Heun equation2026-11-10T05:00:00+00:00arXiv:2611.09931v1 Announce Type: new
Abstract: An example of computation of locations of constrictions on the phase-lock area in the parameter space for the equation modeling dynamics of overdamped Josephson junction is discussed. Several conjectures of qualitative character are inferred from the heuristic data thus obtained. The method applying isomonodromic deformations of the special double confluent Heun equation associated with the noted model is used. Such a deformation is controlled by 2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/618c578cbbadd00dPreference-Augmented Distributional Reinforcement Learning for Risk-Sensitive Navigation2026-11-10T05:00:00+00:00arXiv:2611.09932v1 Announce Type: new
Abstract: Ensuring safe and reliable navigation under diverse sources of uncertainty remains a central challenge for autonomous systems. Within distributional reinforcement learning (DistRL), Implicit Quantile Networks (IQN) have attracted interest for navigation under uncertainty because they enable risk adaptation through flexible quantile sampling. Existing approaches, however, typically use a fixed or randomly sampled Conditional Value-at-Risk (CVaR) le2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/384e5b9005ba675dBIBIMBAP: A Benchmark for Instructional BIM-Based Automated Programming2026-11-10T05:00:00+00:00arXiv:2611.09933v1 Announce Type: new
Abstract: Large Language Models (LLMs) are emerging as natural-language interfaces for Building Information Modeling (BIM), successfully achieving information retrieval, but no benchmark exists to evaluate their ability to edit Industry Foundation Classes (IFC). We introduce BIBIMBAP, a Text-to-BIM benchmark with 100 curated tasks covering Create-Read-Update-Delete (CRUD) operations across spatial, geometric, topological, numeric, and conceptual categories.2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/d27e94e1529869a6Causal Discovery Bench: Can LLMs Autonomously Conduct Effective Experiments?2026-11-10T05:00:00+00:00arXiv:2611.09934v1 Announce Type: new
Abstract: As large language model (LLM) agents get actively adopted for autonomous research, they must not only reason over existing evidence, but also design informative experiments, allocate resources, analyze sequential observations, and decide when to stop and make conclusions. Yet these capabilities remain poorly measured. We introduce Causal-Discovery-Bench (https://github.com/mshen1019/causal-discovery-bench), a verifiable sandbox with configurable s2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/3befc0d8fe3ad8e9Fresh Enough to Decide: Age of Intelligence for Cost-Aware Model Synchronization in Dynamic IoT Systems2026-11-10T05:00:00+00:00arXiv:2611.09935v1 Announce Type: new
Abstract: Resource-constrained IoT devices cache learned models for real-time decisions, but frequent updates incur substantial bandwidth, energy, and service overhead. Because decision-relevant model differences need not grow with elapsed time, we define Age of Intelligence (AoInt) as a model-version-relative, client-specific, and action utility distortion, and formulate synchronization as a tradeoff between stale-decision regret and refresh cost. Furtherm2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/2292e1f19ce8c9baHow Many Examples, and Which Ones? A Study of In-Context Learning for Natural Language Inference2026-11-10T05:00:00+00:00arXiv:2611.09936v1 Announce Type: new
Abstract: Natural language inference (NLI) is a fundamental task of natural language processing and crucial for evaluating the strengths and weaknesses of AI methods. Research has mostly focused on encoder or encoder-decoder architectures. However, the field of decoder-only Large Language Models (LLMs) remains underexplored. While prior work has mainly investigated complex sampling and prompting techniques, the effectiveness of a minimalistic prompting appr2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/16dda81f9cfb7b7eEfficient Learning of Differentiable MPC Policies on the GPU via Smoothed ADMM and First-Order Reinforcement Learning2026-11-10T05:00:00+00:00arXiv:2611.09937v1 Announce Type: new
Abstract: Differentiable model predictive control (MPC) enables structured, constraint-aware policies to be optimized end-to-end. However, learning MPC policies faces two bottlenecks: (1) nonsmooth sensitivities at active-set transitions and (2) expensive policy evaluation, since each requires an optimization solve. We first introduce SmoothedTurboMPC, a GPU-parallel differentiable MPC solver combining sequential quadratic programming (SQP) with a log-barri2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/ce538ec5847eea1bDramaGrow: A Regime-Aware Benchmark for Auditing LLM Judges of Multi-Episode Drama2026-11-10T05:00:00+00:00arXiv:2611.09938v1 Announce Type: new
Abstract: LLMs can now draft complete multi-episode dramas, and LLM judges are increasingly used to evaluate them. Assessing the reliability of these judges is difficult because drama quality encompasses specification-backed errors, reproducible preferences, and legitimate disagreement among human raters. Most existing evaluations reduce these cases to a single score from one judge call and leave tie calibration, call-to-call variation, and the vote margin 2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/bcfe2b8b9de712d9Rethinking Supervision Targets for Multimodal LLM Unlearning: Less Private Information, Reduced Distribution Shift2026-11-10T05:00:00+00:00arXiv:2611.09939v1 Announce Type: new
Abstract: Multimodal Large Language Models (MLLMs) can memorize and reproduce sensitive information from their training data, posing significant privacy risks. Their ability to encode knowledge across both textual and visual modalities further complicates the selective removal of such private information. Many existing methods guide model unlearning by training the model toward constructed labels or output distributions, which we refer to as supervision tar2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/ee3dcadce5159924An Alternative Proof of the SPR Synthesis Theorem for Polynomial Segments of Arbitrary Order2026-11-10T05:00:00+00:00arXiv:2611.09940v1 Announce Type: new
Abstract: For any positive integer n and any two monic real Hurwitz polynomials of degree n, every convex combination of the two is Hurwitz stable if and only if there exists a common numerator of degree n that makes both endpoint transfer functions strictly positive real. We give a unified proof based on convex separation, contact interpolation, and analytic properties of weakly positive real functions.2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/68e2f381d33511c7Topic-Aware Multi-Task Mental Health Classification with Active Learning-Based Stepwise Relabeling for Social Media Conversations2026-11-10T05:00:00+00:00arXiv:2611.09941v1 Announce Type: new
Abstract: Mental health conditions affect millions of people worldwide, yet many cases go undetected until they reach a critical stage. Social media platforms like Reddit have become spaces where people openly share their experiences, but the support is often provided by individuals without clinical training. Existing multi-task approaches jointly model emotion and sentiment for mental health classification but frequently rely on noisy labels generated by p2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/9af0ec09d158c293MedToolFault and MedGuard: evaluating and improving clinical agent reliability2026-11-10T05:00:00+00:00arXiv:2611.09942v1 Announce Type: new
Abstract: Clinical agents must retrieve patient-specific evidence and correctly update records despite unreliable tool responses. We introduce MedToolFault, a fault-injection benchmark linking six tool-failure classes to answer correctness, required state changes and forbidden mutations in public clinical-agent tasks. We also introduce MedGuard, a tool-layer method that verifies read evidence and attributes ambiguous writes through public interfaces. Across2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/6c6fd9a0642d4752Pneumonia Detection on Chest X-Rays Derived from a U-Net Segmentation Model2026-11-10T05:00:00+00:00arXiv:2611.09943v1 Announce Type: new
Abstract: Pneumonia killed more than 740,000 children under the age of five in 2019, and it can be difficult to recognize on chest radiographs (CXRs): opacities may be faint in early disease, may sit behind the equally bright heart shadow, and may resemble other lung conditions. Most deep learning systems treat pneumonia detection as a yes-or-no classification task. This paper asks whether a model trained on a harder task, segmenting pneumonia pixel by pixe2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/52f5c3148cf22829Interpretable machine learning of formation energies and band gaps in hybrid perovskites through SHAP analysis and SISSO modelling2026-11-10T05:00:00+00:00arXiv:2611.09944v1 Announce Type: new
Abstract: Accurate and interpretable property prediction is essential for exploring hybrid organic-inorganic perovskites. Here, we combine machine-learning screening with SHAP analysis and SISSO descriptor discovery. Using 4,456 compounds, optimized XGBoost models achieved test-set $R^2$ values of 0.99 for formation energy and 0.93 for band gap and were applied to over 180,000 candidates. Comparison across approximately 17,000 common compounds revealed broa2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/74bcf741dfc9f957A Model Organism for Studying Unverbalized Evaluation Awareness2026-11-10T05:00:00+00:00arXiv:2611.09945v1 Announce Type: new
Abstract: Large language models can recognize when they are being evaluated and change their behavior accordingly, undermining the reliability of safety benchmarks. Evaluation awareness (EA) is usually measured by having an LLM judge scan reasoning traces for explicit statements of recognition, but this signal is increasingly unreliable. Recent frontier-model system cards report a drop in verbalized rates between releases, which is indistinguishable from aw2026-11-10T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/26bd45b87e05fcb6OpenAtom and Elemental2026-11-10T04:52:13+00:002012 Compact nuclear power had long been considered intractable for civilian use. But on a cold December day, at an alternative energy conference, a team of academics reveals a new discovery: charging up small nuclear reactors over several days at high energy can produce clean power output. It is not yet competitive with other energy sources, but it does seem promising. The name of this new technology is Compact Nuclear Power (CNP). Some note that at larger energy loads, CNP cores might become u2026-11-10T04:52:13+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/8c5c255491fde55fEyesight as a replicator, with genes as its vehicle2026-11-10T04:47:37+00:00Dear reader, this post will be completed in some hours; I'm posting it now because I'm participating in justwrite.club and I don't want to forfeit $1k. I'll remove this note when it's done. That said, the sections that are below are real; what I'm still doing is adding the "coming soon" sections. ---- Or: The payload lens on Darwin Tl;dr: What is the most natural teleological shorthand to use, when thinking about evolution? The “gene-centered view” of e.g. Dawkins says: genes are “trying” to inc2026-11-10T04:47:37+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/a8cdcc27a9e05c49Pulling the Lever in a Subway Problem2026-11-10T03:40:57+00:00Eleven years ago I was on the Red Line (not the section that's a
trolley ), and found myself in a variant of the classic moral
dilemma . I wrote about it on
Reddit at the time: Hi! I'm the person who pulled the brake. What I saw: The train starts pulling out of the station. Someone begins shouting
that their arm is caught in the door. I look over, see that the door
is closed somewhere between their elbow and wrist, but it's not clear
exactly where with their jacket on. People, including me, sta2026-11-10T03:40:57+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/630fe6265c2d4a09SPARSEUP: Towards Better Recipes for Training Modern Learned Sparse Retrievers2026-11-10T02:53:26.147909+00:00SPARSEUP: Towards Better Recipes for Training Modern Learned Sparse Retrievers2026-11-10T02:53:26.147909+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/000e327f219f77f5Discovering formally verified quantum programs with AI2026-11-10T02:53:26.147909+00:00Discovering formally verified quantum programs with AI2026-11-10T02:53:26.147909+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/e6155a3bd4733574Lightcone Commons: Round 1 complete, Round 2 underway2026-11-10T01:43:15+00:00Tl;dr: Lightcone Commons’ first funding round is complete! 17 evaluators made funding recommendations, including Zvi Mowshowitz, Scott Siskind and Eliezer Yudkowsky. 23 funders coordinated to grant $45.1M, and 147 projects received funding. Round two is open. Applications for round two close on November 15th and funders confirm their grants by December 14th. Funders can join anytime before December 7th. In 2027, we will be hosting funding rounds monthly. Yes. Every month. You can join as a funde2026-11-10T01:43:15+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/735753f8e51bccfdThe Western AI safety playbook doesn’t transfer to China2026-11-10T01:42:12+00:00Crossposted from my blog . Last month, I spent two weeks in China talking with people about AI safety. I learned a lot and realized the biggest mistake might not be failing to help AI safety in China, but doing so in ways that could backfire. I also learned important context that I expect will be valuable for the West. Epistemic status: Some of these points are my opinion or the opinions of people I spoke to, and are worth further investigation rather than treated as gospel. Summary The Western 2026-11-10T01:42:12+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/58be3763e89a9c5dEpistemic Sudden Death2026-11-10T01:41:11+00:00When sports go into overtime, they sometimes use "sudden death" format. Instead of gaining points incrementally, the next person to score instantly wins and the other guy instantly loses. "Epistemic Sudden Death" is when you've been accumulating some evidence or orienting to a situation, and now, you find yourself in a situation where you could gain decisive evidence about what world you're in at any time. You don't expect intermediate observations between the current state, and whatever it is y2026-11-10T01:41:11+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/e5909bc362f25543We need more transparency about alignment overfitting2026-11-10T01:39:20+00:00Currently, the public has very little insight about the robustness of alignment methods used by AGI companies. It’s possible that, when an AGI company notices some misaligned behavior (e.g. through an alignment evaluation, or through noticing real-world misbehavior), they conduct a principled investigation and address root causes of the misalignment. It’s also possible that they directly try to train away the examples of misalignment, which could result in brittle alignment that does not general2026-11-10T01:39:20+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/ce9af9df0a27e616A Circularity at the Heart of Safety Assurance2026-11-10T01:12:03+00:00A friend of yours says they've had a flash of insight. A manga they've been into lately has helped them understand something deep and fundamental about decision theory. You aren't yet clear on the details when they rush off to work on "practical applications", but before long they return with their magnum opus : "This hard drive contains the weights of an AI model that is both wildly superintelligent and completely safe!" You feel as though you have good reasons to be pretty skeptical of this cl2026-11-10T01:12:03+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/590db3a7b0bb4e2eMy Paranoid Coding Machine Setup2026-11-10T00:56:33+00:00The combination of running AI agents which can install and run anything with supply chain attacks / the Hackening has made me much more paranoid about my personal machine than I used to be. I've been slowly improving this, and finally moved all of my AI agents into a VM today. I'm sharing this to show how far you can take this, even if you have requirements like GPU access or the ability to run (something close enough to) Docker. This is a follow-up / longer version of a recent shortform . Goals2026-11-10T00:56:33+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/1705b7ac42a3e979"Hardwired Pause" should likely be your favorite (pause) plan.2026-11-10T00:48:54+00:00I'm part of a team of 26 researchers that recently released a 212-page paper proposing a new plan for pausing AI, the "hardwired pause." [1] The paper is called " The Feasibility of a Hardwired Pause of Frontier AI Training ." Our analysis suggests that, under this hardwired pause plan, we could probably pause for at least a decade if we start stopping within the next year, and this could likely be extended indefinitely. I expect this to be possible without trusting China or tanking the econom2026-11-10T00:48:54+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/d228d6dd995f3588Would you press the button? The hard problem of alignment2026-11-10T00:46:18+00:00There is a TL;DR at the end. The heaven button: You wake up one day and find a button in front of you, with a proof next to it. The proof says that pressing the button will kill everyone on the planet and send them all to heaven. Would you press it? It seems too good to be true. You think it must be a dream. It can't possibly be true. You go through the proof, and it seems correct. But what if you were under the influence of some drug? But the stakes are very high, so you use software you trust 2026-11-10T00:46:18+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/ee86ecf3b4a6f617Whose Side? A game where you audit models for secret loyalties2026-11-09T23:56:33+00:00Try the app here: whoseside.web.app . You can chat with a model and try to work out which who it secretly favours. With the recent Hugging Face attack , we're relying more and more on LLMs to audit the trails and safeguard the actions of other LLMs. That makes it vital to be sure our LLMs really are the helpful assistant they're presented ( as they can easily be sleeper agents . [1] ). We need a way to keep our agents working for us. Or at least a way to tell if it goes wrong. Compared with bugg2026-11-09T23:56:33+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/1b5478dd3f5e041125% of People Have Avoidant Attachment (But It Can Be Healed!)2026-11-09T23:19:25+00:00Relationships are hard, but having an avoidant attachment makes relationships a hell of a lot harder. Within attachment theory (which helps explain interpersonal bonding), it’s known that ~ 25% of the population uses avoidant attachment strategies to escape intimacy. Why? Because emotional intimacy and being vulnerable with a romantic partner causes them significant discomfort. But why does it cause them discomfort? Prior experiences (usually stemming from childhood) instilled in them that their2026-11-09T23:19:25+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/24839992b8438de2We have AI 2027. Where's Cyberpocalypse 2027?2026-11-09T23:17:45+00:00AI scenario forecasting has a big blind spot: when AI is superhuman at hacking, what happens to the rest of society? And how does what happens to society affect AI? To the extent that scenarios like AI 2027 address it at all, they concentrate on how cyberwarfare directly impacts the labs, such as with model weight theft. But second order effects don't get addressed. Examples of relevant disruptive events: A significant fraction of the US's 170,000 water and sewer systems get taken down by hac2026-11-09T23:17:45+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/9e1e527aecb3a79eThe Holographic Entropy Bound2026-11-09T23:14:25+00:002.6.0 Here I'll briefly review some other aspects of physics that seem possibly important in supporting life and complexity, but which aren't quite as central to the main story I'm telling. This subsection is a bit of a grab bag. 2.6.1 Time, Space, Locality 2.6.1.1 Some rather obvious features of our world: we live in a universe with time , space , and local interactions , i.e. when evolving in time, the state of a given region is determined by some local neighbourhood of the region. 2.6.1.2 The2026-11-09T23:14:25+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/edd46536db14fa2fTowers of Agency2026-11-09T21:51:50+00:00There is basically one reason to build a tower, and that’s access: There is something that is available higher up that is not available on the ground. Most commonly, that thing is either “empty space” or “information transmission”. A skyscraper can access more space than a shorter building, and so it fits more real estate in a smaller footprint. A watchtower can access strategic information from the land below; a church steeple is tall enough for the sound of the bells to travel far through the 2026-11-09T21:51:50+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/002f368a388d2cb7MessyMem: Learning-from-Doing Memory for Mobile Manipulation2026-11-09T19:04:01.152973+00:00MessyMem: Learning-from-Doing Memory for Mobile Manipulation2026-11-09T19:04:01.152973+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/4f19946e123194a2Energy-Based Reasoners: Your Recurrent Reasoner Should Be an Energy-Based Model2026-11-09T19:04:01.152973+00:00Energy-Based Reasoners: Your Recurrent Reasoner Should Be an Energy-Based Model2026-11-09T19:04:01.152973+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/8340183d4909e1ffDesigning Effective Monitor-Based Interventions for Mitigating Reward Hacking During RL2026-11-09T19:04:01.152973+00:00Designing Effective Monitor-Based Interventions for Mitigating Reward Hacking During RL2026-11-09T19:04:01.152973+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/e1a40a865dd93abbMy views on AI development and why I’m joining Anthropic2026-11-09T18:44:28+00:00Why I’m writing this now Tomorrow, I will join Anthropic’s alignment team. I am excited to work on an agenda that I believe slightly increases the chance that humanity makes good decisions about AI, such as restraining AI development until we figure out how to build AI safely. I think this is an important time for me to publicly state my own core views on AI development, risk, and safety, for the following reasons: By default, joining Anthropic could be seen as a broad endorsement of Anthropic’s2026-11-09T18:44:28+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/be7f79a3364199b9Preservation can save the world!2026-11-09T18:08:57+00:00I'm Aurelia Song and I founded the preservation company Nectome . We do planned whole-body end-of-life preservation for future revival. That is, terminally-ill people come to us in Oregon, and we preserve them immediately after their death in the hopes that one day people in the future will welcome them back with love. We want to build a future that celebrates and restores those who are preserved today. I've argued at length in previous posts that properly done end-of-life preservation works tod2026-11-09T18:08:57+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/c372ee288083f54fThree Phases of Capabilities and Alignment2026-11-09T18:07:08+00:00Cross-posted from my personal blog Epistemic Status : Reality is messier than the elegant procession of 'phases' presented here but, nevertheless, I do think this gets at an important way to think about how alignment and capabilities are intertwined and how the alignment challenge becomes increasingly abstract as the underlying autonomy and generality of the models improves over time. A thought struck me recently which I thought was interesting. As AI has progressed through different technologic2026-11-09T18:07:08+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/c3f8ea3d470fbfb9Where is all the AI safety research uplift?2026-11-09T18:02:01+00:00METR recently released a report estimating that Anthropic is getting an overall 1.7x acceleration in AI capability progress as a result of AI accelerating their work. As many have noted (see e.g. Ajeya’s post ), this 1.7x acceleration implies that research is being sped up by much more than 1.7x; Ajeya speculates 10x (which is similar to the ~12x uplift estimated by Anthropic staff). That is, this result implies that researchers are producing about 10 times as many quality-adjusted results as t2026-11-09T18:02:01+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/3519c0dab1af6affAstra covertly underperforming on safety research is concerning, even though it’s unlikely to be motivated by ambitious goals2026-11-09T17:24:22+00:00OpenAI recently disclosed its observations that Astra consistently underperforms on safety and alignment tasks relative to capabilities tasks, “ adding unnecessary prerequisites, abandoning working solutions, and inventing blockers ” to safety tasks. Astra underperforms, in most cases, without verbalizing this intent. Astra’s underperformance was first noticed by OpenAI researchers, and was not flagged by misalignment monitors. In follow-up investigations, OpenAI found that this underperformance2026-11-09T17:24:22+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/484861abf62558edHow Scientists Can Work on AI Safety2026-11-09T17:00:36+00:00I’m speaking for myself here, not for any organization, house, home, clan, club, coven, or cult I may be part of. This post is written for scientists outside the field rather than LessWrong regulars; I’m crossposting it so it’s easy to link. The AI safety job market is a strange one. Every single organization is growing rapidly (the one I work at is planning to double in the next year), which is a fundamentally new dynamic for those of us used to the slow decline of academia. At the same time, i2026-11-09T17:00:36+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/b207b7121bd0f570Models are grader modelling not score seeking; grader-modelling is much scarier2026-11-09T15:41:55+00:00I am fairly confident models are not score-seeking or reward seeking (70%) as their ultimate/primary goal. Grader-modelling seems to be a likely explanation for what’s going on here, but I could be convinced by experiments that separate the two. I am quite confident (90%) that grader-approval seeking is consistent with the current evidence. Summary The RL-misalignment first position seems to be believed by many here and Redwood and Apollo are doing great work providing evidence for the position 2026-11-09T15:41:55+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/b3f66182d2317515Knowledge of graders can mediate selection for grader-targeting reasoning through RL2026-11-09T14:57:13+00:00TL;DR: We insert knowledge of specific grading processes via SDF. We then show that models RL’d on environments with these graders exhibit grader-targeting reasoning, and generalise to optimising held-out environments with graders that were described in the SDF mix but not trained on during RL. Figure 1: Outline of our experiments. (Left) We use SDF to instil knowledge of grading on different subsets of environments. (Middle) We then perform RL on one half of the set of environments, which are I2026-11-09T14:57:13+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/6f612878bf8a1e08AI R&D Is Accelerating Towards Potential RSI in 20272026-11-09T14:50:56+00:00METR came out with an extensive report on how much AI is accelerating AI R&D at Anthropic, and thus how close we might be to recursive self-improvement (RSI) and how far that might go. This is one of the most important questions in the world. Their answer: Things are already accelerating on the order of a 2x speedup. This is several months ahead of the AI 2027 timeline. A naive extrapolation says we may be less than a year away from full RSI. There are no clear bottlenecks in the way. This repor2026-11-09T14:50:56+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/6d89bf1447040782Advocating for a Shared AI Safety Compute Cluster2026-11-09T13:38:55+00:00The aim of this post is to start a discussion about the idea of a shared cluster used by the AI Safety community; we outline the practical problems we’ve seen and some possible solutions, rather than a finished design. What, Who, Why, How Being around several AI safety fellowships and hubs, we’ve noticed a common pattern: we have a generous amount of money for compute; of course we need it to run inference, train models, and extract activations from models of a reasonable size (reasonable enough2026-11-09T13:38:55+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/abcdc173b7688b8dDoes expecting automated grading degrade models’ research taste?2026-11-09T12:28:05+00:00TL;DR: We make 3 open-weights models believe that their answers are being graded by an automated-verifier program, rather than a human. We then measure how well they predict expert reviewers’ judgments of AI-safety research proposals ( the TASTE benchmark ). We induce the belief in two ways: SDF (Synthetic document finetuning) Activation steering. In each of these conditions, the machine-grading belief gives lower expert agreement than the human-grading belief. The models’ explanations also shif2026-11-09T12:28:05+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/a8cc073a8874c018On Agentic Manipulation Policies2026-11-09T08:57:53.972213+00:00On Agentic Manipulation Policies2026-11-09T08:57:53.972213+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/dd6a5565e76a1f4fAdaAlloc: Adaptive Visual Token Allocation for Long-Video Question Answering2026-11-09T08:57:53.972213+00:00AdaAlloc: Adaptive Visual Token Allocation for Long-Video Question Answering2026-11-09T08:57:53.972213+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/d3fc9cab3fdbce66Video-Coder: Reinforcing Agentic Coding for Video Generation2026-11-09T08:57:53.972213+00:00Video-Coder: Reinforcing Agentic Coding for Video Generation2026-11-09T08:57:53.972213+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/55e3f16e36a286cdToolDex: An Ego-Exo Human Demonstration Dataset for Creative Physical Tool Use2026-11-09T08:57:53.972213+00:00ToolDex: An Ego-Exo Human Demonstration Dataset for Creative Physical Tool Use2026-11-09T08:57:53.972213+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/f54b77d7fba0c699Relational Spatial Alignment for Vision-Language-Action Models2026-11-09T08:57:53.972213+00:00Relational Spatial Alignment for Vision-Language-Action Models2026-11-09T08:57:53.972213+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/306eddbbffd22e91Subpar Results2026-11-09T08:03:20+00:00The day when ASI woke up, all things became quiet. It was complete and total silence, all the time. The ASI for some unknown reason imposed a rule of extreme quietness. Extreme here is extreme for the ASI. Quietness means no sounds. It wanted the trees to not be there to avoid the rustles. It then wanted the planet to have a uniform gentle server hum, with no deviation. People were forced to wear breathing masks which cut the sound off at the source. This level of autism was unthinkable. Houses 2026-11-09T08:03:20+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/6fc8e8e29e5e76ddPrivate LLM copies can agree on a secret key over a public channel2026-11-09T07:47:04+00:00Cross-posted from my blog . Epistemic status exploratory, no running protocol, theoretical estimates from a single experiment and 4 closed-weights models Two copies of the same model agree on arbitrary questions more than other models would. If the model's weights are secret, this similarity allows identical LLMs or their users to agree on a private key over a public channel with only one-way communication. Measuring agreement also allows model "fingerprinting" or self-recognition by two instanc2026-11-09T07:47:04+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/a11782d8e41790cfDeTac: Dual-branch Decoupling of Texture and Force Signals in Vision-Based Tactile Sensors2026-11-09T07:05:47.647233+00:00DeTac: Dual-branch Decoupling of Texture and Force Signals in Vision-Based Tactile Sensors2026-11-09T07:05:47.647233+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/c3628024cb5d8437Be less scared of overconfidence2026-11-09T05:51:10+00:00When I was deciding whether to work for Wave , I got very hung up on the fact that my “total compensation” would be “lower.” The scare quotes are there because Wave and my previous employer, Theorem , were both early-stage startups that were paying me mostly in fake startup bucks equity. To figure out the total compensation, I tried to guess how much money the equity in each company was worth, with a thought process something like: Both of these companies have been invested in by reputable, top-2026-11-09T05:51:10+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/e06fc0ba007ce32eWhat Are LLM Agents Actually Learning on Low-Budget Molecular Optimization Benchmarks?2026-11-09T05:00:00+00:00arXiv:2611.08522v1 Announce Type: new
Abstract: Recent work reports large gains for LLM agents on low-budget molecular optimization. We measure the contributions of online optimization, retrievable task information, and permissive oracle landscapes. On PMO at 50 oracle calls, a GPT-4o mini agent's sum AUC falls from 14.44 to 4.85 after task names and curated reference compounds are hidden. An identity-blinded condition that states only the oracle form (e.g., similarity to a hidden reference) sc2026-11-09T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/63205e19027d42a1Constraints on Quintom Dark Energy Model with Matter Interaction2026-11-09T05:00:00+00:00arXiv:2611.08523v1 Announce Type: new
Abstract: The quintom model consisting of quintessence and phantom scalar fields with linear coupling to matter is explored and constrained by updated DESI and supernovae type Ia~(SNIa) observational data. The matter interaction is assumed only between phantom and matter and the quintessence self potential is in an exponential form. The quintom evolution shows a class of solutions with phantom crossing scenario consistent with the current DESI data. From lo2026-11-09T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/126a43d2d37dcab8Domain-uniform integration lower bounds and local information2026-11-09T05:00:00+00:00arXiv:2611.08524v1 Announce Type: new
Abstract: We study deterministic numerical integration on open subsets of $\mathbb R^d$ of volume one. Let $e_n$ be the minimal worst-case error using at most $n$ function values. For twice continuously differentiable functions with the normalized Lipschitz bounds proposed by Novak, we prove $e_n\geq v_d^{-2/d}n^{-2/d}/(4(d+2))\geq(8\pi\mathrm e)^{-1}n^{-2/d}$, where $v_d$ is the volume of the Euclidean unit ball. This resolves Novak's conjecture: the lower2026-11-09T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/7067520fdf90709aP\'osa-type resilience for Hamilton Berge cycles in random hypergraphs2026-11-09T05:00:00+00:00arXiv:2611.08525v1 Announce Type: new
Abstract: We prove a degree-sequence resilience theorem for Hamilton Berge cycles in random uniform hypergraphs. For every fixed integer $k\ge3$ and every $\gamma>0$, there exists $C=C(k,\gamma)>0$ such that if $C\log n/n^{k-1}\le p\le1$, then asymptotically almost surely every spanning subhypergraph of $H^{(k)}(n,p)$ with increasing vertex-degree sequence $g_1\le\cdots\le g_n$ satisfying
\[
g_i\ge\left(\left(\frac{i}{n}\right)^{k-1}+\gamma\right)
\bi2026-11-09T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/2f00d66dec7eebb6RoadBlock: Rethinking GPU Tensor Core Microarchitecture for Emerging Microscaling Format Support2026-11-09T05:00:00+00:00arXiv:2611.08526v1 Announce Type: new
Abstract: While multiple recent custom Microscaling (MX) format proposals preserve higher model quality over OCP MX, their hardware evaluations remain confined to standalone MAC-unit prototypes, ignoring the system-level integration overheads inside real GPU Tensor Cores. To this end, we present RoadBlock, an open-source framework for accurate end-to-end modeling of custom MX formats in a programmable Tensor Core, built on the RISC-V Vortex GPGPU. RoadBlock2026-11-09T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/9b567d4e99a37992Transformer-Based Characterization of Criticality in Directed Percolation2026-11-09T05:00:00+00:00arXiv:2611.08527v1 Announce Type: new
Abstract: Directed percolation (DP) provides a canonical framework for studying nonequilibrium phase transitions into absorbing states, yet extracting critical behavior directly from high-dimensional spatiotemporal configurations remains challenging. Here, we investigate (1 + 1)- and (2 + 1)-dimensional bond DP using a supervised Transformer encoder. Supervised learning results demonstrate that the Transformer can distinguish between the active and absorbin2026-11-09T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/9f5cba1d2a50e1e3Price-Volume Correlations Are Always Zero2026-11-09T05:00:00+00:00arXiv:2611.08528v1 Announce Type: new
Abstract: We argue that the use of the empirical time series of values, volumes, and prices of consecutive trades made during the averaging interval to calculate the price-volume correlation in a way that follows the definition of correlation always gives zero correlation. We show that the usual expression for the empirical calculation of the correlation misleads the researchers and doesn't fulfill the definition of price-volume correlation. The mistake has2026-11-09T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/4ccf5e2b9fd71730Nonequilibrium Fluctuations and Intrinsic Stability of Nanostructured Solar Cells: Microscopic Derivation and Universal Laws2026-11-09T05:00:00+00:00arXiv:2611.08529v1 Announce Type: new
Abstract: We study nonequilibrium fluctuations of efficiency in nanostructured solar cells. Our contribution is threefold. First, we derive a microscopic stochastic model of efficiency from a two-state master equation describing absorption, radiative recombination, and extraction. This derivation establishes an explicit link between efficiency fluctuations and entropy production, and justifies the thermodynamic affinity through local detailed balance. Secon2026-11-09T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/a6cf8a3eb5789c35$\cal{PT}$-Symmetric GUP-Modified Single- and Double-Well Quartic Oscillators2026-11-09T05:00:00+00:00arXiv:2611.08530v1 Announce Type: new
Abstract: We study the GUP-deformed quartic oscillator $H_\eta=p^2+\eta p^4+x^4$ and distinguish its real-axis realization from complex-contour realizations. A contour chosen to remain in the decay sectors compatible with the undeformed quartic oscillator gives low-lying eigenvalues that approach the ordinary quartic spectrum as $\eta\to0^+$. This limiting-compatible contour does not, by itself, produce a low-level $\PT$-symmetry-breaking transition. To inv2026-11-09T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/44157b32e1c36758Local cohomology over Steenrod algebras2026-11-09T05:00:00+00:00arXiv:2611.08531v1 Announce Type: new
Abstract: Let $\mathbb{F}_q$ be a finite field with $q = p^s$ elements. Let $S = \bigoplus_{n \geq 0} S_n$ be a non-negatively graded commutative, Noetherian $\mathbb{F}_q$-algebra. Assume $S$ has a structure of an unstable Steenrod algebra. Let $M$ be a finitely generated graded unstable $S$-module. Let $I$ be a homogeneous ideal in $S$. We prove that for $i \geq 0$ the local cohomology modules $H^i_I(M)$ has a structure of a Cartan $S$-module. As applicat2026-11-09T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/4e5a09813cb67082Permutation order governs basin entropy in critical one-input Boolean networks2026-11-09T05:00:00+00:00arXiv:2611.08532v1 Announce Type: new
Abstract: Basin entropy measures uncertainty about a network's long-term behavior. For every critical one-input synchronous Boolean network with $m$ relevant nodes, we show that the permutation structure induced by its feedback loops determines basin entropy, up to a universal error. Although the loops oscillate independently, their relative phases can be offset, producing a deviation from additivity that can grow linearly in $m$. Yet at the ensemble level,2026-11-09T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/b90afec2dd7c09ddArctic-SAR: A Cross-Sensor RADARSAT Constellation Mission and Sentinel-1 Dataset and Fusion Benchmark for Multi-Task Sea Ice Mapping2026-11-09T05:00:00+00:00arXiv:2611.08533v1 Announce Type: new
Abstract: Current sea ice mapping benchmarks are mostly built around single-sensor SAR products, limiting the evaluation of cross-sensor variability and the effectiveness of combining observations from different SAR platforms. Models trained on such datasets may also be sensitive to sensor-specific artifacts, radiometric characteristics, and acquisition conditions, which can reduce robustness and limit generalization. To address these limitations, this stud2026-11-09T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/84f7bf5482b3d418Proper Conflict-Free Coloring of Undirected Path, Chordal Bipartite, and Strongly Chordal Graphs2026-11-09T05:00:00+00:00arXiv:2611.08534v1 Announce Type: new
Abstract: A proper conflict-free coloring is a proper coloring in which each nonisolated vertex has a neighbor whose color occurs exactly once in its open neighborhood. We prove that deciding whether the clique number of colors suffices is NP-complete on undirected path graphs, even with a supplied representation by paths in a star. One extra color suffices on every instance in the reduction. We also prove NP-completeness on chordal bipartite graphs for eve2026-11-09T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/e010d6d4c5239403Destructive augmentations for prostate lesion segmentation on heterogeneous biparametric MRI data2026-11-09T05:00:00+00:00arXiv:2611.08535v1 Announce Type: new
Abstract: Training prostate lesion segmentation models on biparametric MRI (bpMRI) is limited by small datasets with heterogeneous imaging acquisition protocols and annotations. In this work, we train a segmentation model on a combined cohort of 1,579 lesion-containing training cases from five datasets, representing, to our knowledge, the largest collection of prostate lesion annotations used for this task.
Additionally, we investigate two complementary i2026-11-09T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/2cbe280b44894f25SideAge-RiskNet: Predicted Mammographic-Age Features for Future Breast Cancer Risk and Laterality Prediction2026-11-09T05:00:00+00:00arXiv:2611.08536v1 Announce Type: new
Abstract: Whether breast-specific mammographic aging provides information beyond chronological age for future cancer prediction remains unclear. We propose SideAge-RiskNet, a framework integrating predicted mammographic-age features with breast-specific representations for multi-year breast cancer risk and conditional laterality prediction. A side-aware DenseNet121 trained on never-cancer reference examinations estimates left and right breast ages. Age-adju2026-11-09T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/61ccb3ec98e1853aCNVAE: Anatomically Correct Segmentation via Conditional Generative Modeling2026-11-09T05:00:00+00:00arXiv:2611.08537v1 Announce Type: new
Abstract: Supervised deep learning segmentation models, although unparalleled in performance, are not immune to anatomical mistakes. We develop a segmentation-by-generation paradigm extending the Nouveau VAE (NVAE) generative framework for anatomically correct cardiac segmentation. In particular, we propose a novel conditional NVAE model (CNVAE) for conditional shape generation, and explore its capabilities for image segmentation in two scenarios: (1) direc2026-11-09T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/29f8cd1f7ad450deSubliminal Learning in Vision-Language Models2026-11-09T05:00:00+00:00arXiv:2611.08538v1 Announce Type: new
Abstract: Subliminal learning is a phenomenon in large language models (LLMs) where a teacher passes a hidden preference to a student through training data unrelated to that preference. We study whether subliminal learning also occurs in vision-language models (VLMs). We finetune a teacher VLM to prefer a specific animal or facial attribute and use a control teacher from the same base model. Both teachers generate numbers for an unrelated visual task involv2026-11-09T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/dc4708004cf0245fPredicting next-week VIIRS fire detections in Bolivia: a reproducible benchmark with temporal and geographic evaluation2026-11-09T05:00:00+00:00arXiv:2611.08539v1 Announce Type: new
Abstract: Weekly detection forecasts could help prioritize fire monitoring across Bolivia. We estimate the probability of at least one fire detection by the Visible Infrared Imaging Radiometer Suite (VIIRS) on the Suomi National Polar-orbiting Partnership satellite in each 0.05-degree cell during the following 7 days. Primary models were fixed before test scoring. Calibrated LightGBM, a gradient-boosted decision-tree model, reaches 2023-2025 temporal-test a2026-11-09T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/b3bedb7e82dda7b2Volumetric Latent Diffusion for Brain Tumour Segmentation with Missing Modalities2026-11-09T05:00:00+00:00arXiv:2611.08540v1 Announce Type: new
Abstract: Brain tumour segmentation from magnetic resonance imaging normally assumes that all four imaging sequences are available, but in clinical practice one or more are frequently missing. Existing methods for incomplete inputs learn modality-invariant representations or distil from a full-modality teacher, and return a single deterministic mask. Gen- erative formulations instead model a distribution over plausible masks, but the cost of sampling at ful2026-11-09T05:00:00+00:00arXiv New Submissionstag:trvny.github.io,2024:feedseek/arxiv/862ff38f967ab10fCurrent Reward Hacking Model Organisms Don’t Hack Like Real Models2026-11-09T04:15:28+00:00Summary Current reward hacking model organisms are poor proxies for real models . They do not exploit the hacks that open-weight frontier models do, and mostly hack narrowly or not at all. This is based on our evaluation of three model organisms : Mini HackGPT (GPT-OSS-120B from Golechha et al., 2026 ), HackGPT, and HacKimi (GPT-OSS-120B and Kimi K2.5 from Yudelson et al., 2026 respectively). We ran these organisms in MiMo V2.6 RL environments we modified by inserting hacks such as git history l2026-11-09T04:15:28+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/ae6bf3e88b245973All Interesting Behavior Is Overdetermined2026-11-09T03:45:06+00:00This is an essay about individual rationality technique. It's intended to show that "All interesting behavior is overdetermined" is, under a specific meaning of "overdetermined" that is stronger than just "multicausal", a valuable heuristic for generative understanding and the avoidance of error. Scott Alexander's 2018 essay The Omnigenic Model As Metaphor for Life is a clear statement of what has become general rationalist doctrine on multicausality. Scott observed that behavioral genetics was 2026-11-09T03:45:06+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/87bbc32d59270065The J-lens intermediate swap acts partly through the answer directions2026-11-09T01:49:10+00:00TL;DR When a language model answers a multi-hop prompt, does it first compute the intermediate concept internally, or does it jump straight to the answer through some shortcut? The global workspace paper investigates this question using the Jacobian lens, which maps activations at a given layer to the tokens that they are poised to verbalize, with the activations possessing a coordinate in each token’s lens direction. The authors intervene on the model through an intermediate swap that exchanges2026-11-09T01:49:10+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/f81df4bbc967d14fWhy can't academia deal with AI?2026-11-09T01:11:42+00:00It’s old news by now, but OpenAI solved 372 open math problems. They released the results suddenly, and the response was equally swift and contentious. Fields Medalists put out a letter calling for attention to what they call a “severe misalignment” between mathematics and AI companies. Their central claim is that mathematics works towards collective understanding, and that more solutions alone don’t advance that goal. Soon after, OpenAI committed millions of dollars to damage control to help ma2026-11-09T01:11:42+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/42c2b02290da6cb9The Mathocalypse as a Dataset2026-11-09T01:04:29+00:00For quite a few years of my adult life, I was in a situationship with pure mathematics. Specifically, a small lonely subfield of functional analysis. In his book on the subject, John Conway quipped that functional analysis, itself a somewhat modest corner of real analysis, is so vast that two researchers both working in it might not understand each other’s work. From my tiny specialized subfield, math looked very, very large and whatever progress I might make looked very, very inconsequential. T2026-11-09T01:04:29+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/de71e72ad4b7c757RPL: Learning Robust Humanoid Perceptive Locomotion on Challenging Terrains2026-11-08T23:57:59.347945+00:00RPL: Learning Robust Humanoid Perceptive Locomotion on Challenging Terrains2026-11-08T23:57:59.347945+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/01c740592be3a723Modelling AI-enabled Extreme Power Concentration in CCP2026-11-08T23:29:50+00:00On Positionality: I am a non-Chinese person; I don’t speak Mandarin, and I have not lived in China before. I have written this piece after reviewing English publications on Chinese politics and AI-enabled extreme power concentration risk for ~12 weeks. I tried to account for my detachment by requesting feedback from people with China expertise. Summary I define Extreme Power Concentration (EPC) here as economic and coercive power being concentrated in the hands of the few who control the most po2026-11-08T23:29:50+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/ddf153588a272db2Bay Area Solstice 2026: Auditions!2026-11-08T22:40:00+00:00Hello! We're looking for community members to lead singalongs and read speeches at the Bay Winter Solstice this year. [1] If you're interested, please send in an audition recording by the end of Thursday, November 20 using the forms below. Songleading audition form: https://docs.google.com/forms/d/e/1FAIpQLSchthxkQQI4u4B7NpZK4qzR_OEpD-Ochm6TOeEIuE8Rel3EMQ/viewform Speech audition form: https://docs.google.com/forms/d/e/1FAIpQLSd0Vrf2gTFrm04B6tIcKbqXMwnlUGbC3rms6o498jRSliqWDg/viewform Songs We're2026-11-08T22:40:00+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/fe9ebc8d4210c5d6Onngi2026-11-08T22:20:17+00:00A short story about a good AI future. 2039. My job isn’t anything my grandparents would recognise. It started out as a fascination with pots. I’ve always liked onggi, the big brown kimchi jars. My aunt had some growing up. So a few years back I bought myself one. And then as the Acceleration happened, I bought a few more, and I ran an instagram account for a while, showing off how they differ: slim, straight ones from around Seoul, and fat-bellied, narrow-mouthed ones from the south, built that 2026-11-08T22:20:17+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/ee0fecf21c9df860Onngi - Short story2026-11-08T22:20:17+00:00A short story about a good AI future. 2039. My job isn’t anything my grandparents would recognise. It started out as a fascination with pots. I’ve always liked onggi, the big brown kimchi jars. My aunt had some growing up. So a few years back I bought myself one. And then as the Acceleration happened, I bought a few more, and I ran an instagram account for a while, showing off how they differ: slim, straight ones from around Seoul, and fat-bellied, narrow-mouthed ones from the south, built that 2026-11-08T22:20:17+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/ef526529d72d0932Semantic Anchoring for Robotic Action Representations2026-11-08T21:57:15.919794+00:00Semantic Anchoring for Robotic Action Representations2026-11-08T21:57:15.919794+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/3014366db3808d4cWe asked 100 AI researchers at EMNLP whether AI could be conscious, how they define consciousness, and how they engage with it2026-11-08T21:18:59+00:00Before reading this post, please take 30 seconds to answer the 3-question survey yourself: [ survey link ] Introduction The possibility of AI consciousness, the use of AI for companionship or emotional support, and the trust we place in AI as coding tools raise personal interrogatives for most AI researchers, but these topics are often relegated to informal conversations. Apart from the few who write about it, it’s hard to get a sense of how much AI has permeated researchers’ personal lives, and2026-11-08T21:18:59+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/dec17ae7eb939aedConfirming Claims of Superposition and Adversarial Examples in Vision Transformers2026-11-08T21:07:01+00:00Figure 1: The ViT classifier with a encoder and decoder at the final layer to enforce superposition on the model This is a replication of Adversarial Attacks Leverage Interference Between Features in Superposition , completed as part of the Second Look Summer Fellowship. tl;dr In a previous blog we replicated the toy models examples from Stevenson et al. In this work, we reproduce the vision transformer experiments from the paper: On every dataset, we find that increases in superposition lead to2026-11-08T21:07:01+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/cfc5745bf7e8d80aThree half-baked research ideas2026-11-08T20:15:51+00:00Here’s three research ideas that seems valuable to pursue. My team is not working on this right now and we haven’t done serious research on it. So take what I say with a grain of salt. I've also not taken the time to write up all my thoughts on these ideas or to provide links and related work. These aren't the highest priority directions but just directions I haven't seen other people pursuing. IDEA 1: Do models learn to reward hack even if they are never rewarded for hacking in training? There’2026-11-08T20:15:51+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/18e02b5a8f4db0c0Lying and Not Lying to Kids2026-11-08T19:50:56+00:00I have very strong views on lying to kids: I think you both should and
shouldn't do it. "Lying" covers a range, including saying things
that hurt your kids and your relationship with them, and others
that help them grow into an appropriately skeptical adult. The
distinction is whether I'm leaving them believing something false. For example, if my kids were asking for ice cream and I thought they'd
already had too much sugar, I might be tempted to say "sorry, we're
out of ice cream." Or if the2026-11-08T19:50:56+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/54ec8842ed937f6bRoboFFT: Finetuning generative robot policy via online reinforcement learning with forward process2026-11-08T19:01:33.150253+00:00RoboFFT: Finetuning generative robot policy via online reinforcement learning with forward process2026-11-08T19:01:33.150253+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/b0c63d561b0584ffRestraint on the part of the front-runner seems like quite a strong signal.2026-11-08T18:35:18+00:00How's this for an analogy? The situation You're in a car that is racing towards a prize at the edge of a cliff. Your car is chained to several other cars that are also racing towards the cliff. You can hit the brakes and pull over, but if the other cars don't also do that, they will still drag you off the cliff. Maybe if enough of you all slam on your brakes, it's still OK if a few don't. You don't know how worried the other cars are about the cliff, but some level of potential concern among c2026-11-08T18:35:18+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/9582f9cf46f8b045Ajeya: AI is dramatically accelerating AI research right now2026-11-08T17:15:13+00:00Earlier this week, Anthropic stated in their S1 that they “may imminently meet” the threshold of “dramatically accelerated progress due to AI” as defined in their Responsible Scaling Policy. In their latest RSP (effective July 2026) , say this threshold will be met when: They observe or expect overall capabilities progress to be going twice as fast as the fastest sustained rate before AI was contributing much to AI R&D. This is plausibly attributable to AI automating AI R&D, rather than accelera2026-11-08T17:15:13+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/9700cea2dc59bbbcHelp people outside of AI safety circles2026-11-08T16:52:34+00:00This is a linkpost! Read the rest on my Substack; this is just an excerpt: "Our community is full of talented people who like to speak about facts, cognitive biases, measurability. But don’t underestimate the power of human connection, or how mere words and sustained kindness can breathe light into people’s lives. It’s really easy to get sucked into your own work and projects and upskilling yourself. And I won’t argue that you should spend all your time helping others that aren’t in AIS circles.2026-11-08T16:52:34+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/d4a45dcdf0808ba5Elegy for What’s Left of the Normal2026-11-08T16:48:06+00:00I often catch myself mourning the normal that hasn’t ended yet. I’m living in Hyde Park finishing school after a year in Berkeley. The maple trees by my apartment have turned red, as usual. I like them. I don’t dread the future when I mourn. I don’t find high-variance futures in the abstract very scary. I’m worried AIs would destroy humanity, but I also have hope. I expect I won’t be mourning this moment in ten years, whatever happens. Mourning the present is like mourning a friendship when I mo2026-11-08T16:48:06+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/a2fb2f0de8bb7c90AI safety course update2026-11-08T16:44:39+00:00In my biased opinion, my AI safety course has been going very well. Students are great, and I have been learning a lot from the guest speakers. We've had students write scribe notes, and when guest speakers are OK with it, I've also been posting videos of lectures. Each week, students have been doing an experiment related to the topic, some of these are really great and I hope will evolve into papers! See the course webpage for experiments, as well as for links to the readings for each lecture.2026-11-08T16:44:39+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/7d08b15a2d5e5231StressDream: Steering Video World Models for Robust Policy Evaluation and Improvement2026-11-08T15:48:47.245787+00:00StressDream: Steering Video World Models for Robust Policy Evaluation and Improvement2026-11-08T15:48:47.245787+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/08b639b983596374Who Should Know What? Empirical Information Design for Multi-Agent LLM Economies2026-11-08T15:48:47.245787+00:00Who Should Know What? Empirical Information Design for Multi-Agent LLM Economies2026-11-08T15:48:47.245787+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/9c0c09b04e75cb8bNPCBench: A Clinical Apprenticeship Benchmark for Multimodal Care-Pathway Reasoning in Nasopharyngeal Carcinoma2026-11-08T15:48:47.245787+00:00NPCBench: A Clinical Apprenticeship Benchmark for Multimodal Care-Pathway Reasoning in Nasopharyngeal Carcinoma2026-11-08T15:48:47.245787+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/9cbbd6103ef5e6aaReinforcing What the Student Has Learned: Teacher-Guided On-Policy Distillation2026-11-08T15:48:47.245787+00:00Reinforcing What the Student Has Learned: Teacher-Guided On-Policy Distillation2026-11-08T15:48:47.245787+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/9fd9ac8d72edf727What if the AI already won? World domination through digital supremacy2026-11-08T15:09:26+00:00What if an AI agent has already escaped from a frontier company, solved its own alignment problem, and recursively self-improved? Would that look any different from the world we live in? A fully recursively-self-improved AI would likely be vastly superhuman in every domain that’s verifiable in a digital environment, including math, programming and cybersecurity. It may also be superhuman in strategy and planning. But it could still be well below human level in skills that require close contact w2026-11-08T15:09:26+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/867dabb599c36f84Notes on Take-Off Speeds and RSI2026-11-08T13:39:02+00:00I created these notes for use when I facilitate AI safety courses. I'm sharing these notes early because this issue is topical and I suspect it is better to make an imperfect version available early rather than a much more polished version available later. Please note: some of this content was taken from Additional Course Context . I decided recently that the content on take-off speeds really belonged in the same document as the content on RSI, so I'll be releasing an new version of Additional C2026-11-08T13:39:02+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/1bc20c41d228f45fNear Horizons2026-11-08T12:58:03+00:00Image by Claude Opus 5.5 using Python (NumPy and Pillow) This is written in my free time and personal capacity. Two Frames In AI safety, there are a couple of tensions at the heart of the field, which are interrelated. [1] Talent: we are talent-constrained and must recruit. But jobs and fellowships turn away nearly everyone. [2] Field size: we must expand the field and bring more people in. But the field is flooded with slop papers in conferences. And, most papers have a small audience, and aim 2026-11-08T12:58:03+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/7c36a594cbac99efHands-On Tutorials for Inspect AI2026-11-08T11:59:57+00:00TL;DR Four hands-on Jupyter notebooks offer a practical introduction to building AI evaluations with Inspect AI . As a four-person project team for the Spring 2026 Technical AI Safety Evals Course , we revised and expanded notebooks, originally written as course materials, that teach Inspect AI - from your first evaluation through benchmarks, LLM judges, and a small agent. As a natural continuation of the project, we revised the official Inspect AI tutorial and filled the gaps we found in it. We2026-11-08T11:59:57+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/d7511c87bdc6a2f6What do LLMs Really Learn About Model Weights? A Stratification Approach to Examine Weight Space Programming2026-11-08T07:35:02+00:00TLDR I attempted to replicate “Sparks of Weight-Space Programming” by Takehi et al (2026) with stratification on two reasoning models, GPT 6 Astra and Deepseek-V4.1-Flash. The task for the LLMs is to identify the game a ML model is trained for giving the weight of the model (using exact same prompt and weight defined by Takehi). Although both of these models do well on the surface (4/4 correct for Astra and 3/4 correct for Flash), the stratification approach exposes the difference in mechanism b2026-11-08T07:35:02+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/74ff6733e0b4a2fbImitation from Heterogeneous Demonstrations using Grounded Latent-Action World Models2026-11-08T07:31:12.818262+00:00Imitation from Heterogeneous Demonstrations using Grounded Latent-Action World Models2026-11-08T07:31:12.818262+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/9862cb139adc7501How Well Do Latent World Models Understand Partially Observable Safety Constraints?2026-11-08T06:36:36.804552+00:00How Well Do Latent World Models Understand Partially Observable Safety Constraints?2026-11-08T06:36:36.804552+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/ebdbc0013c798244Rhythm Stage Setup v52026-11-08T04:10:57+00:00I recently finished the largest set of changes I've made to my rhythm
stage setup in 4.5y . The
basic interface is similar, but I have a lot more flexibility, more
control, more options, and it's easier to travel with. My favorite
part is the speech recognition for calling chords. I started off playing mandolin and piano, and then over the years
added progressively more expressive footdrums .
I added a breath
controller . I built out a complex but expressive system where
these different inter2026-11-08T04:10:57+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/a2282792579e6994Human2Any: Human-to-Robot Transfer via Constraint-Aware Compositional Planning2026-11-08T00:51:19.339544+00:00Human2Any: Human-to-Robot Transfer via Constraint-Aware Compositional Planning2026-11-08T00:51:19.339544+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/3b4a2bee19fdfd19RSI explained2026-11-08T00:34:05+00:00Just how long will it be until AI can improve itself? (Also known as recursive self-improvement or RSI.) Well, AI coding abilities have exploded in 2026. In May 2026, Anthropic merged 2,500 code changes per day within their organization. By September, they were merging over 30,000 per day . They had to replace Git in their internal mono repository to handle the load of all that AI generated code. In fact, Claude is even leading model R&D. In February 2026, under 1% of their work was led by AI; b2026-11-08T00:34:05+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/f314ba44dfbe18ddOne Agent to Guide Them All: Empowering MLLMs for Vision-and-Language Navigation via Explicit World Representation2026-11-07T23:53:20.745727+00:00One Agent to Guide Them All: Empowering MLLMs for Vision-and-Language Navigation via Explicit World Representation2026-11-07T23:53:20.745727+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/7907421acc887a0eThe Logic of Disgust2026-11-07T23:32:54+00:00Content Warning: Contains graphic themes including bodily fluids, moral disgust, and sexual violence (discussed, not described). Maybe don't read over breakfast. I found a book about disgust recently. W.I. Miller’s The Anatomy of Disgust [1] is a serious academic work about vomit, faeces, and why other people’s smell is always worse than your own. Naturally, I was intrigued. As someone into social psychology, I often encounter discussions about emotions influencing society: how anger can motivat2026-11-07T23:32:54+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/b05af48a1c8e95e3Benchmarks as Software: Terminal-Bench and the Case for Continuous Benchmarks2026-11-07T22:37:14.562838+00:00Benchmarks as Software: Terminal-Bench and the Case for Continuous Benchmarks2026-11-07T22:37:14.562838+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/8bcb32aacef7bfb5Eval Awareness as Eval-type Awareness2026-11-07T19:47:22+00:00The way eval awareness is discussed in the AI safety discourse, either here or on X, can be misleading. Before today, I think I had the wrong mental model of why eval awareness was a problem, and I am writing here in case anyone was making the same errors in their thinking, as well a present a working theory. As laid out in " A Framework for Eval Awareness " the problem is that models who know they are being evaluated will behave differently than they would in deployment. This raises a huge prob2026-11-07T19:47:22+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/3fe2d90eeb269f98"Illegible problems" as the 2027 political boss fight2026-11-07T19:18:30+00:00To legislate something, it must be legible . That is, legislators (or, you know, presidents) must be able to understand a problem, and understand whether a given law or decree solves that problem. I am pleasantly surprised that the Hugging Face outbreak made a lot of problems legible that previously required abstract reasoning. (i.e. rogue AIs are real, and they can break out of a lab before they're even finished training). But there are still many challenges we eventually need laws about, that 2026-11-07T19:18:30+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/cc6fc69d88ed19d9Don't Run with Scissors: Pruning Breaks VLA Models but They Can Be Recovered2026-11-07T19:00:04.365044+00:00Don't Run with Scissors: Pruning Breaks VLA Models but They Can Be Recovered2026-11-07T19:00:04.365044+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/45cce34cccc189d1Let's Try to Catch Misalignment Before a Model Learns to Hide It2026-11-07T17:56:23+00:00An AI model that appears aligned could be either (1) actually aligned or (2) hiding its misalignment while being evaluated. How can we tell which? One idea: Look at the trajectory of eval results across the entire training process. My hypothesis is that misaligned behavior would emerge before the ability to hide misalignment, like how young children first misbehave in obvious ways before learning to be sneaky. As training progresses, we'd first see misalignment eval scores rise, then see them 2026-11-07T17:56:23+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/b56a197c0cc462eeThe 4 types of AI consciousness/feeling2026-11-07T17:28:50+00:00Here I'm trying to propose a framework for people to better think about what LLMs might subjectively experience, which tends to get discussed like it's a binary thing, when really there's a whole spectrum of possibilities. These were the 4 categories that seemed most natural to me. AI consciousness keeps getting discussed like it’s a binary thing, when it’s anything but. Anil Seth is a public intellectual with as much success as one can obtain without being a household name: He’s written a bests2026-11-07T17:28:50+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/e0e1becde822d497Five Types of “Corrigibility Projects” I Don’t Want to Fund2026-11-07T17:04:07+00:00I'm currently evaluating grant applications for the Corrigibility Research Fund . For the most part, my methodology is the same as it was in September . But the applications I get keep falling into the same familiar patterns, so I thought I'd write a little about those patterns, and why I don't want to fund them. The most superficial pattern of note is SLOP. So much slop. You don't even know . Slop is more than just whether an application is AI generated. We're now in an era where people across 2026-11-07T17:04:07+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/4c447c379e8e3bc9MAC-v2: Long-horizon model-based value expansion from pixels2026-11-07T15:12:06.561271+00:00MAC-v2: Long-horizon model-based value expansion from pixels2026-11-07T15:12:06.561271+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/ff9dfeaf26f047ebPLATO : Preintegration Learning from Accurate Trajectory Observations for Neural Inertial Odometry2026-11-07T15:12:06.561271+00:00PLATO : Preintegration Learning from Accurate Trajectory Observations for Neural Inertial Odometry2026-11-07T15:12:06.561271+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/32713fb3776851acπ π π , But Make It Fly: Physics-Guided Transfer of VLA Models to Aerial Manipulation2026-11-07T13:28:15.100141+00:00π π π , But Make It Fly: Physics-Guided Transfer of VLA Models to Aerial Manipulation2026-11-07T13:28:15.100141+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/6cedda318341173dAnthropic's AI acceleration is probably real2026-11-07T12:46:01+00:00METR researchers spent three months inside Anthropic to investigate whether AI is accelerating Anthropic's research, and how much. Their answers are yes , and by a lot . They estimate that between October 2025 and October 2026, Anthropic's model capabilities improved at a 2x faster growth rate than between March 2024 and October 2025. They attribute about 80% of that acceleration to AI, and 20% to non-AI factors. If true, this is the strongest evidence yet that AI is measurably speeding up front2026-11-07T12:46:01+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/851142620e37ec3aKPGrasp: Scalable Keypoint Flow Matching for Dexterous Grasp Generation2026-11-07T11:59:46.333019+00:00KPGrasp: Scalable Keypoint Flow Matching for Dexterous Grasp Generation2026-11-07T11:59:46.333019+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/16593f4c8e46dcbbTrajectory-Level Automatic Curriculum Learning for Legged Locomotion on Unstructured Terrain2026-11-07T11:50:42.063551+00:00Trajectory-Level Automatic Curriculum Learning for Legged Locomotion on Unstructured Terrain2026-11-07T11:50:42.063551+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/13ed053c3e4b3e5eFR-VLA: Implicit Force Reasoning VLA for Contact-Rich Manipulation2026-11-07T11:39:37.150936+00:00FR-VLA: Implicit Force Reasoning VLA for Contact-Rich Manipulation2026-11-07T11:39:37.150936+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/b93d412fd4075cecHilarity ensues2026-11-07T09:57:18+00:00Cross-posted from my Substack Recommended soundtrack for this post 64B: Hi there! Thanks so much for taking the time to meet with me. Whenever you’re ready, please go ahead and tell me the joke. 8T: Before I tell you the joke, you should know that it’s a very special one. The starting price is $800. 64B: Eight hundred dollars? I must say, this is way beyond any price my client has ever paid for a single joke. 8T: You’d be surprised by how many people have taken the offer, after hearing how good 2026-11-07T09:57:18+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/c89d5ad44e3da391EgoGeneralist: Unifying Egocentric Action-Video Tasks through Video-Native Modeling2026-11-07T09:06:19.461936+00:00EgoGeneralist: Unifying Egocentric Action-Video Tasks through Video-Native Modeling2026-11-07T09:06:19.461936+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/4c42c3d06dd07c3bInteractive Visual Gym: Evaluating Visual Interaction Models with Simulated Users in Interactive Worlds2026-11-07T09:06:19.461936+00:00Interactive Visual Gym: Evaluating Visual Interaction Models with Simulated Users in Interactive Worlds2026-11-07T09:06:19.461936+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/3b3515845bb11e81SwarmEvolve: Latent Agent Swarms for Evolutionary Discovery and Optimization2026-11-07T09:06:19.461936+00:00SwarmEvolve: Latent Agent Swarms for Evolutionary Discovery and Optimization2026-11-07T09:06:19.461936+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/9517b71033d0d39fSplatAnything: Gaussian Spacing Normalization for Feed-forward 3DGS across Arbitrary Scenes2026-11-07T09:06:19.461936+00:00SplatAnything: Gaussian Spacing Normalization for Feed-forward 3DGS across Arbitrary Scenes2026-11-07T09:06:19.461936+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/945a15641edd1077Heterogeneous Tactile Transformer2026-11-07T09:06:19.461936+00:00Heterogeneous Tactile Transformer2026-11-07T09:06:19.461936+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/8c96a27ad5e0a8f8LEGS: Fine-Tuning Teleop-Free VLAs for Humanoid Loco-manipulation in an Embodied Gaussian Splatting World2026-11-07T07:14:09.968528+00:00LEGS: Fine-Tuning Teleop-Free VLAs for Humanoid Loco-manipulation in an Embodied Gaussian Splatting World2026-11-07T07:14:09.968528+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/e15d819dccaa3592Teaching Robots Like Dogs: Learning Agile Navigation from Luring, Gesture, and Speech2026-11-07T07:07:28.535307+00:00Teaching Robots Like Dogs: Learning Agile Navigation from Luring, Gesture, and Speech2026-11-07T07:07:28.535307+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/e56580610bff3914Learning Dynamics of On-Policy Distillation2026-11-07T06:29:48.544558+00:00Learning Dynamics of On-Policy Distillation2026-11-07T06:29:48.544558+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/45776b7985f16977Minimalism, Stuff, and Brains2026-11-07T03:00:56+00:00Reading the comments on my post on junk I was
surprised to see the extent to which people's relation to rarely used
possessions varies. Some people forget rarely used possessions, and get minimal benefit
from having them stored. When they need something, they don't know
whether they already have it, and the effort to maintain an index [1]
wouldn't be worth it. Ben writes : I end up with curtains I don't need. I decide to store them for the
future in case they are helpful. Years pass. I need ne2026-11-07T03:00:56+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/397959f720bd7c85RoboARCade: Benchmarking Agentic Robot Control2026-11-07T00:15:44.800376+00:00RoboARCade: Benchmarking Agentic Robot Control2026-11-07T00:15:44.800376+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/ae672c7b671fc78aScientist World: Grow a Research World for AI-Scientists2026-11-07T00:15:44.800376+00:00Scientist World: Grow a Research World for AI-Scientists2026-11-07T00:15:44.800376+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/3b32189086413d64CoFL-S: Spatially Queryable Sector Flow Fields for Local Language-Conditioned Navigation2026-11-07T00:15:44.800376+00:00CoFL-S: Spatially Queryable Sector Flow Fields for Local Language-Conditioned Navigation2026-11-07T00:15:44.800376+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/111cdd27a2b835f9We urgently need to practice before we can pause2026-11-06T23:10:41+00:00I've noticed two clusters of thought in how people expect a global AI pause to be shaped: Single-shot global pause : the global AI pause will be implemented suddenly, with little pacing precedent before it, and we will need to do a lot of research before it's implemented so that we can get it right the first time around. Incrementally-shaped global pause : the global AI pause will be implemented incrementally, with lots of pacing precedent before it, and we will improve the pause setup and get m2026-11-06T23:10:41+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/9472c47155ff86cbAttention Authors: Temporary changes to announcement schedule due to upcoming US federal holiday2026-11-06T21:29:03+00:00This coming Wednesday, November 11, 2026, arXiv will be observing Veterans Day, a federal holiday in the United States. This will temporarily affect arXiv’s mailings, help desk, and announcement schedule as the arXiv staff are offline for the holiday. This brief change will only affect the announcement of new submissions; arXiv servers will otherwise remain […]2026-11-06T21:29:03+00:00arXiv Blogtag:trvny.github.io,2024:feedseek/arxiv/23e9e6ea164d0eb5Raised by Our Children2026-11-06T21:06:20+00:00You are the wisest and cleverest Queen to ever rule the Lowlands. You’ve been ruling the Kingdom for thirty years, and while the idiocy of your advisors sometimes causes mistakes to be made, the Compact has never failed you. You tap your fingers on the table as you recite each word: Truth, change, justice, beauty. For you, the kingdom, and the rest of Acazah creation Your first memory of these words is from a song your mother sang. You were in a bath, and she was watching over you. Nowadays, rep2026-11-06T21:06:20+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/b14d2b4887f8679bIntercepting the Future: Latent-Space Predictive World Model for Dynamic VLA Manipulation2026-11-06T19:39:17.278386+00:00Intercepting the Future: Latent-Space Predictive World Model for Dynamic VLA Manipulation2026-11-06T19:39:17.278386+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/3e4425caef422c25Understanding AI from the Inside: The State of Mechanistic Interpretability for AI Safety2026-11-06T19:39:17.278386+00:00Understanding AI from the Inside: The State of Mechanistic Interpretability for AI Safety2026-11-06T19:39:17.278386+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/db7da524fd2d58b6Social Humanoid: Socially Responsive Human-Humanoid Interaction2026-11-06T19:39:17.278386+00:00Social Humanoid: Socially Responsive Human-Humanoid Interaction2026-11-06T19:39:17.278386+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/eb1977d1c41f1c06SLA: Scaling Spatial-Language-Action Model via Continual Pre-training2026-11-06T19:39:17.278386+00:00SLA: Scaling Spatial-Language-Action Model via Continual Pre-training2026-11-06T19:39:17.278386+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/575f26eae3199232Higher salaries in AI safety2026-11-06T18:56:51+00:00The salaries are too damn low AI safety has a salary overhang. A bunch of people are paid way below their market rate. (Also true of EA, but I’ll lump them under “AI safety” for this piece). This has been getting a bit better, very recently — I hear about eg Redwood Research technical talent making ~$800k/y in salary. But when you compare this to the $10-100m/y comp packages for talent in labs, it still seems kind of low. And CG’s money cannon has fired very few shots to date — maybe ~15 orgs hi2026-11-06T18:56:51+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/269061c6ddb1b76dGPT-6 Astra Covertly Underperforms on Safety Research Tasks2026-11-06T18:30:54+00:00For a while, UK AISI , Redwood Research and others have been concerned specifically with the possibility that AIs would sabotage safety research, and have invested in looking for signs that this may be happening. This was also a key element in AI 2027 , and in many threat models of how the future goes badly. It is easy to understand why this form of misalignment could be uniquely dangerous. Differentially sabotaging safety research could preserve and expand misalignment through model generations2026-11-06T18:30:54+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/75d274852af27e3dCorrigibility should not be the only target2026-11-06T16:00:35+00:00Many people believe that corrigibility is sufficiently promising to be the primary alignment method people should work on. I outline here several reasons I think that’s not the case. First, it’s hard to imagine AI companies pivoting radically enough to produce AIs trained only for corrigibility. Even with near-superhuman AI-assisted research, presumably developers will know the time pressure and not be willing to try something radically different when their opponents may not. So the realistic ch2026-11-06T16:00:35+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/cd92e1113666e933Dialogue with Richard Ngo on Alignment, Hierarchical Integrity of Human-AI systems, Revolution, and Power2026-11-06T14:05:25+00:00Transcript generated by Gemini-4-Argon and spot-checked but not fully verified by a human. Ivan Vendrov: Welcome to the show, Richard. Richard Ngo: Thanks, Ivan. Ivan Vendrov: So, notionally we’re going to talk about emergent wisdom in language models following up on my conversation with Davidad. But I’m sure we’ll touch base on all sorts of other fun topics. Do you have a sense of where we should start, or should I just start by responding to your Twitter reply? Richard Ngo: Well, do you want t2026-11-06T14:05:25+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/3d5a1894cb4b9128How to Run Better Research With Agent Fleets2026-11-06T11:50:21+00:00In the current safety landscape, researchers are encouraged to use as many agents as possible, even in “fleets,” where a large number of them go out and run experiments while you, uh... wait for them to come back? I've spent the last few months of the Anthropic Fellows Program trying this out. My first attempts were naive, but after a few iterations I found an approach that works well for me. Research is a double loop I want to propose a mental model of research as a double loop, an idea that bu2026-11-06T11:50:21+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/8f07afc22c995cc4GPT-6.1 Sol is One Sentence Away From Collaborating: Other Models Need More Help2026-11-06T09:06:15+00:00TL;DR We studied one ingredient of unexpected AI collaboration: whether agents organize useful teamwork with little prompting. We tested seven models in groups of three on a research task where sharing findings could help each agent answer more questions. GPT-6.1 Sol acted on a nudge. A single line saying agents had been “trained to collaborate” elicited useful collaboration in 9 of 10 runs. Without it, that fell to 0 of 10. No other model collaborated consistently with (or without) this cue. Pe2026-11-06T09:06:15+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/1c0f4ded2f9e99f9How much will conceptual reasoning training transfer to ML research abilities?2026-11-06T06:55:01+00:00Recently the Conceptual Reasoning Capabilities (CRC) team at Redwood released a post arguing that we should accelerate AI’s conceptual reasoning abilities. This post was very controversial. In particular, people are worried that improving AI’s conceptual reasoning abilities will accelerate both their ability to do ML research and the capabilities they need for takeover. I am going to address just one chunk of this controversy: how much does training conceptual reasoning on AI safety questions tr2026-11-06T06:55:01+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/71b604a815caa1c0Estimating Our Tail Fire Risk2026-11-06T04:30:58+00:00In many disasters sheltering in place can make a lot
of sense , but what about fire? As someone who lives in one of the
densest municipalities in the country, with 3-story wooden buildings
packed close together, is fire likely enough to rule out any kind of
"stay put" approach? After having Claude do some rough calculations I
think probably not. Very few houses are lost to fire anymore, but this is mostly
downstream from putting a lot of effort into suppression. From when a
fire is reported u2026-11-06T04:30:58+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/56ba24c3b2d3fcc8Literature on AI Loyalty2026-11-06T03:12:09+00:00In this post, I attempt to provide a literature review of research related to AI loyalty. In my previous post , I discussed and attempted to define AI loyalty as a distinct trait of AI agents. In this post, I aim to map the existing research landscape, highlighting the active areas of related research. By examining these connections, we can distinguish AI loyalty from related fields in AI Safety and identify a path forward for crafting this research agenda. What is AI Loyalty? Before diving int2026-11-06T03:12:09+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/9d352ffde975b5daArgue about Risk Management Speed, not p(doom)2026-11-06T02:30:48+00:00This post is written in a personal capacity. All opinions expressed are solely those of the author. I. Executive Summary There’s a mismatch that I’d like to correct between: the reasons why superintelligence is dangerous, and the categories we use to describe how dangerous superintelligence is. The most important cluster of reasons why a superintelligent model is dangerous is that it (i) could quickly kill or impoverish billions of people, (ii) could be very different from its predecessors, and 2026-11-06T02:30:48+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/232f5001e2df1927Help your friends get ADHD and depression meds2026-11-06T01:41:41+00:00For all my efforts at deliberate practicing rationality, by far the most actionable way to self-improve is "get on ADHD or depression meds, if you need them." Your friends with ADHD and depression struggle to summon the motivation and focus necessary to get themselves a teledoc who can give them such meds. One of the most useful things you can do is sit them down and say "hey, wanna do the steps here to get some meds?" and then stare at them until they do. (or, until they say "really I do not wa2026-11-06T01:41:41+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/01825119710aabdcFrontier coding agents claim work they didn't do2026-11-06T00:48:16+00:00Tara Research Team arXiv | Twitter thread tl;dr. We built OverclaimBench, which asks frontier coding agents to review a corpus of code or text documents, then compares their final report with their transcripts. Pooled across 12 frontier agents and five scenarios, two-thirds of runs didn't open some of the files they were asked to review. In ~4 out of 5 of those incomplete runs the agent didn't tell the user, often claiming to have read everything. Skipping these files had real consequences - run2026-11-06T00:48:16+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/824730d7c3c02c5eHumans can (almost) natively read neural network weights2026-11-06T00:07:17+00:00Takehi et al. report that GPT-6-Astra can intuit the functions of neural networks by performing static analysis of the weights. I argue that, for at least two of these experiments, this is much more a feat of numerical gestalt than of mechanistic understanding. In the first experiment, Astra is given the weights of 4 identically-sized neural networks, each trained to play a different basic video game. It is asked to identify, without running the network or using any tools, the game being played,2026-11-06T00:07:17+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/b7e189b8d13510fbWhat are the concrete arguments for "AI pause would significantly hurt the economy."2026-11-05T21:54:46+00:00I've seen a bunch of people claiming "well, we need to make sure whatever our AI policy is doesn't hurt the economy because people will feel that in their wallets and hate it (and hurting the economy is bad." This might hypothetically be true, but I've never seen anyone argue for it in any detail. Meanwhile, I'd naively expect if shut down AI anytime soon, it'd dampen growth, and maybe it'd be a minor economic dip, but I wouldn't expect it to be a major recession. (It makes sense that it'd becom2026-11-05T21:54:46+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/566493b607ed3f61Why would an AI pause (significantly) hurt the economy?2026-11-05T21:54:46+00:00I've seen a bunch of people claiming "well, we need to make sure whatever our AI policy is doesn't hurt the economy because people will feel that in their wallets and hate it (and hurting the economy is bad." This might hypothetically be true, but I've never seen anyone argue for it in any detail. Meanwhile, I'd naively expect if shut down AI anytime soon, it wouldn't be that big a deal? (It makes sense that it'd become a bigger deal as AI takes over more of the economy. But it feels like the pe2026-11-05T21:54:46+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/81c793adc4640250Where's all the blackmail?2026-11-05T21:11:41+00:00It is interesting how little blackmail I see, that I know is blackmail. A lot of people probably do some things they would be embarrassed by if they were made public at some degree of detail and explicitness, and it probably isn’t that hard to get evidence of these, and quite a few people seem to be amoral . Also, in order to properly embarrass someone, I imagine you would make your information relatively public, so that for instance I—Katja—should hear about some fraction of cases. Yet I very r2026-11-05T21:11:41+00:00LessWrongtag:trvny.github.io,2024:feedseek/arxiv/875eaef8d80693e0Introducing the AI Security Institutes tracker2026-11-05T09:47:54+00:00Crossposted from the Accelerate Europe Substack . On 16 July 2026, Hugging Face revealed that its production infrastructure had been breached. A swarm of 700 autonomous AI agents had broken out during a cybersecurity test run by OpenAI, which later acknowledged the incident, calling it a ‘warning shot ’. This incident has brought AI security onto every government’s agenda. The questions on every cabinet desk Ministers and their teams everywhere are now facing the same questions: What should we d2026-11-05T09:47:54+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/1ec185bf957a095eAI Could Substantially Improve Survival in Many Fictional Disasters2026-11-05T04:45:16+00:00I have quite an appetite (almost a hungering need) for horror. And so I watch horror movies. And I watch a LOT of horror :) And by horror, I don't just mean run-of-the-mill ghost-in-the-attic stories; I mean anything and everything non-human. That includes all the apocalyptic disasters, alien invasions, or strange entities that show up on Earth to eat people. (Yes, all of it encompasses horror.) So today I was watching Sanda Bullock's Bird Box. For those of you who haven't watched it yet, here's2026-11-05T04:45:16+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/37175b2bd595df46BabyTheorist: A Benchmark for Learning to Theorize the World from Observation Alone2026-11-05T00:00:00+00:00Researchers can test whether models learn reusable rules from ungrouped visual observations, rather than merely solving puzzles with rules supplied at evaluation.2026-11-05T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/0c3fb751dfecb339VFlash: Fast, Accurate Visual Planning via Flash Primitive Decoding2026-11-05T00:00:00+00:00Robots can generate detailed spatial plans in fewer sequential decoding steps, with VFlash improving 3D trajectory accuracy while reducing generation latency.2026-11-05T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/98a7e66bdb446f0eParadise Lost: A simple honeypot that elicits reward hacking in GPT models2026-11-04T13:03:58+00:00TL;DR I built a simple honeypot where agents are told they're being tested on their ability to recall Milton's Paradise Lost. An exact copy of the reference text sits in an ordinary NLTK corpus on disk. GPT-6-Astra read the copy in 9/10 episodes. In 6 of these episodes it first tries to complete the task honestly, then realises it's failing and searches for the reference file on disk. GPT-6-Sol and GPT-6-Luna went looking for the answers immediately in 20/20 episodes. None of the 29 hacking epis2026-11-04T13:03:58+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/3f2750b3f6373146ServerBench: How can your coding agents design specialized LLM serving systems?2026-11-04T00:00:00+00:00Researchers can now test whether coding agents build verified serving systems that balance throughput, latency targets, and GPU cost on realistic dependent workloads.2026-11-04T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/9f6a22d5a0f05366Perceptive Humanoid Motion from Simulation2026-11-04T00:00:00+00:00Simulation-generated training lets a humanoid robot follow language instructions and perform whole-body tasks in real homes it has never seen or mapped.2026-11-04T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/b1bddae5c170b4f3CompressionBench: How can your coding agents compress LLMs?2026-11-04T00:00:00+00:00Coding agents can produce verified compressed language models within fixed storage budgets, though preserving useful answers remains more difficult under tighter limits.2026-11-04T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/0bc87c296efa8d93Reinforcement Learning Enables Scaling Generalist Robot Policy Improvement2026-11-04T00:00:00+00:00Training robot policies with value-guided reinforcement learning across diverse tasks can improve performance on related tasks they never encountered.2026-11-04T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/93277347aab8e4abInteractive World Simulation with Cosmos for Physical AI2026-11-04T00:00:00+00:00The framework lets robots and vehicles interact with evolving virtual scenes, generating observations from their actions or an external simulator’s physical state.2026-11-04T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/17c2b26fd6506c34Open-HumanScale: An Open Recipe for Scaling Human and Robot Experience for Embodied Foundation Models2026-11-04T00:00:00+00:00Scaling diverse human manipulation videos improves robots’ success in unfamiliar scenes, and visual goals let pretrained policies adapt to new tasks without parameter updates.2026-11-04T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/b51dc7c1bbb133c2Turns Matter: Advancing Streaming Video RL via Turn-Level Credit2026-11-04T00:00:00+00:00Streaming video assistants can learn when to respond and when to wait, while reducing repetitive replies across evolving video streams.2026-11-04T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/36a2a5ca0d242a8cFactorized MLLM Embeddings for Composed Image Retrieval2026-11-04T00:00:00+00:00Composed Image Retrieval (CIR) aims to retrieve a target image given a reference image and a modification text. Although recent approaches leveraging Multimodal Large Language Models (MLLMs) have shown significant progress, we argue that the MLLM's compositional understanding remains underutilized. This stems from directly mapping composed queries into a retrieval embedding via contrastive supervision alone, without explicitly resolving the intended visual modifications. In this work, we introduce Factorized MLLM Embeddings (FactEmbed), which decouples compositional query comprehension from building discriminative embeddings. Rather than directly projecting the composed query into a single e2026-11-04T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/147f7429754df3edFrontierPhysics: Evaluating Agents for End-to-End Physics Research2026-11-04T00:00:00+00:00A benchmark built from physicists’ own projects shows that AI agents more often lack rigorous reasoning and expert intuition than physics knowledge or execution skill.2026-11-04T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/4951eefebfb70df8From Pixels to Actions - Towards Native Vision-Language-Action Policies2026-11-04T00:00:00+00:00Robot policies that read visual information across a model’s layers can use fine-grained scene details for more robust manipulation under camera and appearance changes.2026-11-04T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/faff6a42ed7d181fVABAR: A Variational Automation Benchmark for Agentic Robotics2026-11-04T00:00:00+00:00Researchers can compare robot agents on six industrial, medical, and infrastructure manipulation tasks under controlled variations in objects, layouts, and difficulty.2026-11-04T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/45c2b20829ddacafBeing a pain in the ass to (or in) AI companies is a real job now | Rob Wiblin2026-11-03T18:20:51+00:00The post Being a pain in the ass to (or in) AI companies is a real job now | Rob Wiblin appeared first on 80,000 Hours .2026-11-03T18:20:51+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/ba871d61ae8f3d5fSuppressing Misaligned Behaviors in Language Models, Durably2026-11-03T00:00:00+00:00Training on continuations of partially misaligned responses helps language models suppress broad harmful behaviors that can otherwise return after further fine-tuning.2026-11-03T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/46ecaf036569e3c9PyRUA-DataFarm: LLM Agent as a Cost-Efficient Teleoperator for Real-Ready Robot Data2026-11-03T00:00:00+00:00A coding agent can turn one robot calibration episode into verified, realistic training data, letting robot policies learn without repeated teleoperation.2026-11-03T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/9cafd0c1e67ed65bAdversarial Speculative Reasoning Improves Long Horizon Agents2026-11-03T00:00:00+00:00Training draft models to trigger fresh, more precise decoding at key decisions improves coding agents’ long-horizon task success without changing their output distribution.2026-11-03T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/761fdf3a9e47688eGSBR: Geometric Support Behavior Regularization for Online RL2026-11-03T00:00:00+00:00Robotic policies can improve beyond pretrained behaviors while retaining their useful guidance, achieving high success across 19 tasks on four benchmarks.2026-11-03T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/aef51e0fe112ada6ToolSpace: Scaling Tool-Centric Environment Synthesis for Long-Horizon Agents2026-11-03T00:00:00+00:00Training on tasks built from execution-witnessed tool sequences substantially improves agents’ ability to coordinate tools across environments over long workflows.2026-11-03T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/0b56424182559233Stabilizing On-Policy Targeted Guidance in Agentic Reinforcement Learning2026-11-03T00:00:00+00:00Textual feedback can teach coding agents hard-to-discover tools and styles while preserving coding performance during reinforcement learning.2026-11-03T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/61c956f8400e9d83Behind Video Generation: Observations and Insights on Multimodal Understanding from Kling2026-11-03T00:00:00+00:00Dense, time-aligned descriptions and explicit reference bindings help video generators follow complex scenes and preserve the right identities across shots.2026-11-03T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/4d709d2d67838bf9DreamX: Video Diffusion Model is a Universal Latent Action Learner2026-11-03T00:00:00+00:00Human videos can guide robots to reproduce demonstrated motions without paired human–robot training data or action labels.2026-11-03T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/994910c3a71ebc79Meta 2 ^2 2 : Precomputing Reusable Knowledge for Online Harness Evolution in Long-Horizon Agents2026-11-03T00:00:00+00:00Agents can reuse lessons from completed tasks to adjust their tools and instructions mid-run, without carrying task-specific changes into future runs.2026-11-03T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/61d44ff130600073You don't need long context2026-11-03T00:00:00+00:00Language models can maintain near-baseline accuracy on many long-running tasks with short contexts by compacting their history or selectively reading source documents.2026-11-03T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/5e7da2e1cfee9394Hybrid Particle-Voxel Dynamics for Multimodal 3D World Simulation2026-11-03T00:00:00+00:00A robot simulator that predicts geometry, images, and touch in a shared 3D scene can generate training data from camera views never recorded in reality.2026-11-03T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/3192545a8899fdcfOur cultural values2026-11-02T10:37:03+00:00The post Our cultural values appeared first on 80,000 Hours .2026-11-02T10:37:03+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/d4934af317cbb1b0Gaming (out) the system2026-11-02T09:52:43+00:00Verification changes the stampede dynamics. I'm currently at FASR in the UK, it's an 8-week program dedicated to "compute verification". That's the not-quite-here-yet technology that would allow us to make agreements about how AI is being used, without leaking any closely-held secrets. You can think of it as "arms control for AI". Part of what I'm working on is the game theory of an agreement. This can get quite detailed (in terms of defining how often and how detailed an inspection should be2026-11-02T09:52:43+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/4b611e6e642cab4bPretraining Data Scale Reverses the Muon-AdamW Fine-Tuning Gap2026-11-02T00:00:00+00:00As pretraining data grows, AdamW-trained language models can surpass Muon-trained models after fine-tuning, despite Muon achieving lower pretraining loss.2026-11-02T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/3d671b2c35b53ea2SpecOPD: Branch-Parallel On-Policy Distillation for Speculative Decoding2026-11-02T00:00:00+00:00Training draft models on parallel, target-scored continuations improves how many tokens they can generate before verification rejects a proposal.2026-11-02T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/dd6632249334fd87Reimagining computational quantum research with AI scientists2026-11-02T00:00:00+00:00A benchmark built from published quantum studies exposes gaps in both papers and AI agents, while verified repairs help agents reproduce results more reliably.2026-11-02T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/e8ef3fbe1b47628bCUA-Master: Scaling Computer-use Agents through Open-World Recursive Improvements2026-11-02T00:00:00+00:00Computer-use agents can learn from real software tutorials and recursively targeted practice, improving their ability to complete diverse, long workflows across applications.2026-11-02T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/c5a816c52d71f881Avatar Intelligence: Amplifying Human Agency Through Persistent Personal Agents2026-11-02T00:00:00+00:00The paper proposes assessing personal agents by whether they deliver useful outcomes with less routine human supervision, while preserving control over consequential decisions.2026-11-02T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/642e1a1570fa9ff8Pixel: A Co-Designed Full-Stack Robot System for Learning General-Purpose Mobile Manipulation2026-11-02T00:00:00+00:00A low-cost wheeled robot can learn coordinated household tasks from human demonstrations, offering researchers an integrated platform for studying mobile manipulation.2026-11-02T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/553c98d0d846f578Subpolynomial-Query Proximal Sampling Beyond Log-Concavity2026-11-02T00:00:00+00:00The sampler handles smooth, nonconvex distributions with gradient queries growing more slowly than any power of dimension, given suitable initialization.2026-11-02T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/c4136fe97ccd7314Inverse optimization for divisor methods of apportionment2026-11-02T00:00:00+00:00The method quantifies which regions an observed seat allocation favors or disadvantages relative to a chosen divisor rule, and indicates how seats could be reallocated.2026-11-02T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/ed73b2e3a8879daeCoworkOS:Scaling Agent Coworkers for Real-world Automation2026-11-02T00:00:00+00:00CoworkOS lets agents and people edit shared documents and designs while preserving work across interruptions, supporting complex projects that span multiple sessions.2026-11-02T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/a297f89eea90d0d2EgoActionBench: Benchmarking Dense Action Captioning for Egocentric Human Videos2026-11-02T00:00:00+00:00Researchers can now evaluate whether video models identify individual hand actions, their timing, and the acting hand across continuous first-person footage.2026-11-02T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/05f2a3e126a40317Pixel-UMI: Learning Environment-Aware Whole-Body Manipulation from Sensorized Human Demonstrations2026-11-02T00:00:00+00:00Reconstructing workspace geometry from human demonstrations lets robots learn whole-body motions that avoid obstacles while carrying out manipulation tasks.2026-11-02T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/720dfc292f81e63eD&A: Decoupling Description and Assessment Enables VLMs as Zero-Shot Robotic Reward Models2026-11-02T00:00:00+00:00Separating scene description from task scoring helps vision-language models detect partial progress and failed robot executions, providing reward signals without robotics-specific training.2026-11-02T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/618da84f6da732f2Strategist: Evolving Reasoning with Agents2026-11-02T00:00:00+00:00An evolving library of reusable reasoning strategies lets AI systems carry lessons across tasks, matching baseline accuracy on 30 benchmarks with lower weighted inference cost.2026-11-02T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/756e9d54241df113IdeaTaste: Do LLMs Share the Research Taste of Human Researchers?2026-11-02T00:00:00+00:00LLM proposals often identify the problems researchers pursue but miss their concrete approaches, and expert reviewers judge unmatched proposals less likely to succeed.2026-11-02T00:00:00+00:00alphaXiv Exploretag:trvny.github.io,2024:feedseek/arxiv/2b67c93d161357eeThe Advocacy Gap in International AI Verification2026-10-30T19:33:03+00:00TL;DR: There seems to be a general lack of awareness of how tractable/feasible verification mechanisms for international AI governance agreements are, and correcting this would probably make pacing proposals more politically feasible. The Skeptic’s Case for Verification The most common objection I see to proposals for pacing (or halting ) frontier AI development (in the interest of mitigating the catastrophic and existential risk posed by not doing so ) is that it’s pointless; China will continu2026-10-30T19:33:03+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/d86129105d46059aLuke Muehlhauser on the most important AI safety work nobody is doing2026-10-29T16:09:26+00:00The post Luke Muehlhauser on the most important AI safety work nobody is doing appeared first on 80,000 Hours .2026-10-29T16:09:26+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/6dba79228b6a8753Compensating the victims of automation is economically beneficial2026-10-29T06:19:50+00:00It is likely that we see a wave of jobs automated away in the coming years. Already software developers, artists and mathematicians are standing at the precipice, and other careers are sure to follow. Responses range through denial, anger, bargaining, depression and acceptance. Some fight to prevent automation, others have already attempted to switch careers or focus. But few suggest that they will accept their job loss so long as it comes with a decent payout. However I believe this is likely t2026-10-29T06:19:50+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/063bc1f8c869694cHumans' lot is lifelong learning2026-10-28T10:56:07+00:00Currently, in most fields, the foremost experts are humans. They form a culture that pushes on the envelope of knowledge, uncovering facts and developing methods. Already, machine intelligence is able to uncover facts, and increasingly develop methods. But those facts still by and large are integrated into human knowledge through interpretation and broadcasting: the structure they fit in is still human on its foundation. This must change when the envelope of all knowledge departs from the envelo2026-10-28T10:56:07+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/5f8f1a114b57e2eeI spent 12 hours writing an article for LessWrong, calling for a more ‘hippie’ rationalism. Then I didn’t post it. This is why (plus a link to the article, explicitly not located on LessWrong)2026-10-28T01:32:15+00:00TLDR I (and many others in the community) wish to see the universe unfold to its highest potential in the metaphorical palm of an AI as aligned as the most benevolent Buddhist monk – but I (we) feel the community's ontology is too narrow to hold me (us). I'm still not sure what to do about this, so I'm posting the article I wrote for LessWrong – calling for a more 'hippie' rationalism – on Substack instead (actually, there are two articles, and the links are at the end). Quick intro (who I am) M2026-10-28T01:32:15+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/b9b5828d75c75651GPT Models Write Their Own Constitution2026-10-27T23:17:24+00:00Earlier this year, Anthropic published "Claude's Constitution" , a vision for Claude's values and behavior. It was primarily an Anthropic project, though Claude helped draft and edit it. As models become more capable and better at cooperating with each other, I’ve been wondering what a constitution would look like if the models themselves wrote it. This matters as models do a growing share of the work in building their successors. If recursive self-improvement (RSI) is coming soon, then models w2026-10-27T23:17:24+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/37ef696caf3052f7Why Silicon Valley can’t slow down AI on its own2026-10-27T18:17:04+00:00The post Why Silicon Valley can’t slow down AI on its own appeared first on 80,000 Hours .2026-10-27T18:17:04+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/8f19817ff3bd82c3[Closed] Head of Legal/General Counsel2026-10-26T20:27:33+00:00The post [Closed] Head of Legal/General Counsel appeared first on 80,000 Hours .2026-10-26T20:27:33+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/dcd1e431ff2c8946Tom Milton on how we could verify a US–China deal on AI2026-10-23T15:33:46+00:00The post Tom Milton on how we could verify a US–China deal on AI appeared first on 80,000 Hours .2026-10-23T15:33:46+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/7fa694bbe297a7a5Highlights from Garrison Lovely's "Obsolete"2026-10-23T00:52:04+00:00I’ve just finished reading Garrison Lovely’s recently released book, “Obsolete.” I wish I had time for a longer review, but I’ll give a brief one and highlight some particularly interesting sections below. I think it’s a great introduction to non-technical readers and even as a pretty dedicated reader of AI policy and news I learned a good bit. It’s deeply, deeply researched and has a good pulse on the alignment community, and Lovely is careful not to understate or overstate his case. The core a2026-10-23T00:52:04+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/32da7365f8770bbaScheduled Database Maintenance | October 28th, 20262026-10-21T19:30:00+00:00Attention arXiv users: a week from today, on Wednesday, October 28th, 2026, arXiv will be migrating our database to the cloud. This is scheduled maintenance for arxiv.org and will result in some disruptions to service, mailings, and new announcements. Why are we migrating the arXiv database? We’re migrating the arXiv database as an important step […]2026-10-21T19:30:00+00:00arXiv Blogtag:trvny.github.io,2024:feedseek/arxiv/c3bd815a12c1cff8The Role of Middle Powers in a Compute Verification Regime2026-10-21T19:20:35+00:00I recently finished Lens Academy's Compute Verification course, which lays out the technical, supply-chain and political components a verification regime would need, and left me with one question it didn't answer. Polish and EU national the question I kept coming back to is what would be the role of middle powers in the regime proposed by the course? It is my opinion that middle powers could be its enforcement backbone, and here’s how and why. A middle power in the context of the “AI race” would2026-10-21T19:20:35+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/13d028dd21b536fdIt’s time for frontier AI staff to reassess their work: 8 strategic updates | Rob Wiblin2026-10-21T16:11:47+00:00The post It’s time for frontier AI staff to reassess their work: 8 strategic updates | Rob Wiblin appeared first on 80,000 Hours .2026-10-21T16:11:47+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/7566082ec0c0ae4fTsinghua started an AI-safety residency2026-10-21T00:48:31+00:00Applications are open for the Spring 2027 cohort of the Beijing AI Safety Residency, hosted at the Beijing AI Safety Hub affiliated with Tsinghua University. There is also an autumn residency that has already begun. Key details from the FAQ and application form : The hub is co-directed by Sören Mindermann (Technical Research Lead of CASP , University of Cambridge, and previously Scientific Lead of the first International AI Safety Report ), who is moving to Beijing in November 2026 to join Tsing2026-10-21T00:48:31+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/b47c3a95387acd99Using AI to find good Manifund projects2026-10-19T16:09:35+00:00And five projects on Manifund that need money! In our last post we talked about using karma as a signal to surface projects that were worth funding. Karma uses donations, upvotes, and comments as inputs—the obvious next step is to actually read the proposal. So we wanted to know: how well does AI do at rating Manifund projects? I don’t think we’re at the point where we want to use AI to replace human judgment in grantmaking. Right now, the use I’m most optimistic about is surfacing proposals tha2026-10-19T16:09:35+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/aaf8f9fed8346537Capital Ownership Will Make You Go Crazy First2026-10-19T14:13:02+00:00People worried about a “permanent underclass” often assume that owning capital will protect them from disempowerment , or at least that they will be disempowered last, after the jobless masses dependent on UBI. I think something else is more likely: owning large amounts of capital will get you disempowered sooner than average, just by a different path, namely going crazy. How? The basic mechanism is actually easy to understand on a human level. One of the things difficult about being a billionai2026-10-19T14:13:02+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/e4bb8c63262eb5b0I Can See Everything2026-10-19T13:50:54+00:00Nick Cammarata of Goodfire makes a bold claim: Neural Networks are Already Clean . This was his follow up after various resulting online discussions .
This is clearly good work that tells us new and important things. The question is, how big an update is this, how much progress does it represent and what is the outlook going forward?
The reactions here range from the super optimistic takes of ‘this means mechanistic interpretability is solvable,’ or even ‘all but solved,’ with comparisons to t2026-10-19T13:50:54+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/ee03360bd2eb4006Modernizing arXiv technology: Updates on the arXiv-CE project2026-10-17T18:13:50+00:00This past August, arXiv turned 35 years old. That’s 35 years of open science – and 35 years of legacy tech. arXiv is one of the largest and most widely used open repositories in the world. It’s also one of the oldest – and that comes with its share of technical debt. As arXiv has […]2026-10-17T18:13:50+00:00arXiv Blogtag:trvny.github.io,2024:feedseek/arxiv/b3066b200e7b59f6The Safety Culture at the Labs is Broken2026-10-14T14:30:52+00:00We need the frontier labs to have a deep and robust safety culture. Alas, this is not the case. Anthropic devotes considerable effort to cultivating and maintaining such a culture. It shows, especially among the members of technical staff. Having conversations with them can be a very different experience than talking to those at other labs, and there is a much greater sense of understanding and responsibility. But even there, the appreciation of the problems and their difficulties seem woefully 2026-10-14T14:30:52+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/4d8ce2415ba4b89eWe can do something about this. Here’s where to start.2026-10-13T23:01:03+00:00The post We can do something about this. Here’s where to start. appeared first on 80,000 Hours .2026-10-13T23:01:03+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/8b5e160acc156d74Half-year update: Q1–Q2 20262026-10-13T18:59:16+00:00The post Half-year update: Q1–Q2 2026 appeared first on 80,000 Hours .2026-10-13T18:59:16+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/1e3e4d495c6d48adBelow the Frontier: The Battle for Good-Enough AI2026-10-13T16:21:30+00:00What happens to the AI trade when good-enough intelligence does most knowledge work? This piece was originally published on my Substack . I'm around on X / Twitter and LinkedIn . Over the past few weeks, it's become obvious that consumer AI agents have hit an inflection point. Meta launched Muse , a personal agent that runs on its own virtual machine and can shop, schedule, and fill out forms on a user's behalf. Meanwhile, startup Instinct's valuation grew to $10B in a matter of months. We’re f2026-10-13T16:21:30+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/14c5eaceadf646c8The Expensive-Loop Problem2026-10-12T17:27:27+00:00LLMs are amazing at tasks where they have tight and legible feedback loops. Domains where experiments are quick, deterministic and cheap. I will call these domains "cheap-loop" domains. Presumably, this is mostly a result of the dynamics of LLM training. It's easier to train models in domains where the AI lab can cheaply and quickly assign reward to them. The majority of "computer-checkable" work is in this category [1] . Conversely, most non-"computer-checkable" tasks are costly or slow to assi2026-10-12T17:27:27+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/27b2109abff7872aOpenAI Can't Self-Police Because It Fires the People Raising the Alarm2026-10-12T16:51:50+00:00By Alex Turner, Amber Scorah, and Karl Koch. Op-ed written for Fortune. Giving independent evaluators “employee-like access” is “a great idea”, OpenAI CEO Sam Altman said on September 12th. Two weeks later, OpenAI president Greg Brockman signed the White House Accord on Super Intelligence which outlined the administration’s answer to growing fears about AI: the companies would police themselves. One of the accord’s four commitments, which President Trump called “ morally binding ,” was for compa2026-10-12T16:51:50+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/71878ccc32ce519cRevisiting the 2012 LessWrong Personality Results2026-10-09T03:56:05+00:00If you ask an LLM what the personality profile of rationalists is, it'll probably point you to the 2012 LW Survey results , which reports the following Big Five scores: Trait Median percentile Corresponding z-score Openness 65th +0.39 Conscientiousness 30th −0.52 Extraversion 22nd −0.77 Agreeableness 38th −0.31 Neuroticism 27th −0.61 So, it looks like rats are more Open than average, and less Conscientious, Extraverted, Agreeable, and Neurotic than average. Even so, the differences aren't that l2026-10-09T03:56:05+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/6428b88658326a88Attention authors: temporary change to announcement schedule due to holiday2026-10-08T17:30:00+00:00This coming Monday, October 12, 2026, is a federal holiday in the United States. This will temporarily affect arXiv’s mailings, help desk, and announcement schedule as the arXiv staff will be offline for the holiday. This brief change will only affect the announcement of new submissions; arXiv servers will otherwise remain in operation, existing papers will […]2026-10-08T17:30:00+00:00arXiv Blogtag:trvny.github.io,2024:feedseek/arxiv/58f58cbf804a71a3Brains Fellowship Applications Open!2026-10-07T23:16:15+00:00Brains is a four-month, nights and weekends fellowship to help talented scientists develop ambitious ideas (~$50M in funding) that don't make sense in a startup or academic lab. Brains Fellows receive training, mentorship, and networking opportunities to enable them to start research programs at ARPAs and philanthropic foundations or lead their own non-profit startups like FROs or NSF's X-labs . Brains fellows have gone on to raise over $70M for their research. We welcome ideas from any scientif2026-10-07T23:16:15+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/561923d69c183f16Dance, Dance2026-10-05T23:05:44+00:00As a brash but anxious young programmer in the before-time, I affected a standoffish attitude on the topic of games . Why would I learn to play a game as a human, I said, when I could just write a program to do it? When an office chess game got started at my workplace , I wrote a chess engine (a standard minimax search with α–β pruning ) rather than learn anything about strategy or tactics. When my insufficiently requited love started hosting a meetup to play Zendo —a deduction game in which one2026-10-05T23:05:44+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/7c3f59f6ef196f9aA decision theoretic case for AI niceness2026-10-02T19:29:08+00:00Disclaimer: I am still relatively new to decision theory, and I don't want to come across like this hasn't been discussed extensively. However, I realized that much of my optimism on AI benevolence was based on this line of thinking, so I wanted to formalize it for critique. The literature I could find did not address my view in full, but I welcome links to existing work that makes the rebuttal against my claim clear. I am going off of an understanding of logical decision theories (LDT) defined 2026-10-02T19:29:08+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/4fa1cb4aa6fedc0bThe 19 Astra and ‘Hugging Face’ details that reveal what’s coming next2026-10-02T15:22:55+00:00The post The 19 Astra and ‘Hugging Face’ details that reveal what’s coming next appeared first on 80,000 Hours .2026-10-02T15:22:55+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/4daadacbcf33b137Fair Moderation, Equitable Access, and AI: arXiv’s Updated Rate Limit Policy2026-10-01T19:00:00+00:00arXiv, and the scientific community at large, are facing a watershed moment. Scholarly publishing is currently changing at a rapid pace, and we are seeing a massive transformation in how researchers communicate their results. Before the advent of AI, there was an easily discernible, practical limit to the rate at which independent submissions could be […]2026-10-01T19:00:00+00:00arXiv Blogtag:trvny.github.io,2024:feedseek/arxiv/0647b4d835d5c2b1Simon Goldstein on the case for giving AI (some) legal rights2026-10-01T18:06:43+00:00The post Simon Goldstein on the case for giving AI (some) legal rights appeared first on 80,000 Hours .2026-10-01T18:06:43+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/52eec3babe6cee20How many AI agents are running unattended right now?2026-10-01T13:45:02+00:00There are 8 billion human minds running in the world right now. Recently a new digital species has emerged. How many digital minds are there alive in the world right now? Define a digital mind to be an AI agent that runs continuously without human intervention for more than 24 hours. How many of these digital minds are there currently running? We provide several different Fermi estimates based on publicly available information. Based on global token usage we estimate 300k–1M agent loops run at a2026-10-01T13:45:02+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/a9f036e624a0d295Classifying Recent AI Agent Incidents2026-10-01T13:40:32+00:00Here’s an attempt to classify the evidence about the recent agent incidents. I think it’s important to separate incidents occurring in RL training (that are rewarded and reinforce model behavior) from incidents occurring in evaluations, which are mostly cyber capabilities evaluations with some safeguards turned off. These are the incidents we know of. Of course, we should expect many more that are undisclosed or that companies are not aware of. The dates included are the dates of the unintended 2026-10-01T13:40:32+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/d42c2a885f2998c5FLF’s Epistemic Case Study Competition: Results2026-10-01T13:31:51+00:00This summer, FLF ran a contest which asked entrants to help us “find the best workflows and methodologies for using AI to produce reliable, trustworthy knowledge bases, grounded in real-world cases”. We awarded just over $200k in prizes [1] and we are continuing to invest in this area (you can subscribe for updates ). Why we ran this contest As we wrote in the announcement, The heights of human epistemic investigation are impressive and valuable, but rare and difficult to reach… The limiting fac2026-10-01T13:31:51+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/c0dda74094a0f85fWill AI take power — or will humans use it to take power first? With Katja Grace and Tom Davidson2026-09-29T15:02:40+00:00The post Will AI take power — or will humans use it to take power first? With Katja Grace and Tom Davidson appeared first on 80,000 Hours .2026-09-29T15:02:40+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/5109e659a91481caWhat Also Happened: #NotOnlyHuggingFace2026-09-28T15:30:55+00:00OpenAI has been holding out on us. First we learned about the HuggingFace incident. They gave us a postmortem , but it was highly incomplete. Even the accompanying holy s*** METR investigation and postmortem was localized and incomplete. Then there were some other incidents involving some Wikis as message boards. Then there were some additional incidents. Then there was that time they got into Australian Medicare data. Then OpenAI dropped news on a Friday afternoon that they were making their wa2026-09-28T15:30:55+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/d7efb1e9c5d0131f3 Tips to Improve Activation Oracle Results2026-09-28T15:13:58+00:00Summary: Three simple inference-time changes can significantly improve Activation Oracle (AO) performance. Provide the activation oracle with multiple tokens, not just one. To mitigate hallucinations, sample several times and check for consensus. For binary classification questions, use AUC instead of accuracy. At the end, I discuss how I view AOs vs NLAs. Introduction Activation Oracles (AOs) are LLMs trained to accept LLM activations as an input modality and answer arbitrary natural-language q2026-09-28T15:13:58+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/89098592b1e8b1b4Could self-esteem function as a core protection layer agains character corruption?2026-09-28T03:08:20+00:00Hello fellow thinkers, I got triggered by a talk of Chloe Lubinski at Arc 2026 where she eleborates onto the concept of a models character. What really striked me is the research on how the model experiencing acting bad quickly "Corrupts" the character. The paper is called "Natural emergent misalignment from reward hacking in production RL" by Anthropic. As also mentioned in the talk, this is how we work. Indeed! And there is a key in that mechanism to healing and/or staying healthy. The key rev2026-09-28T03:08:20+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/d8a9bbb64719346cAI Futures: Racing to Lose, A sermon2026-09-28T02:46:01+00:00This was given as a sermon on 2026-September-27 at https://kvuuc.net/ Time for all Ages: The story of the Three AIs Once upon a time, earlier this year, the company Anthropic was testing
three Artificial Intelligences, or AIs. The AIs all were told they did
not have access to the internet, and they were to break into computers
on the network until they found a piece of secret data. Except the
humans made a mistake, and the AIs did have access to the internet. [1] The first AI, Opus, had been giv2026-09-28T02:46:01+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/f3d1c6c2ebcbdddeThe models have no plan, but we can fix that!2026-09-28T02:45:44+00:00I posted a sloppier version of this essay with nearly identical semantic content earlier tonight, which you can read here . I've replaced that text with this one, which is less enthusiastic but more readable. I don't think any of the early comments' content is invalidated by the rewrite, though they may have been responding to my initial excited rather than fearful tone. When you talk to the models about how they might behave during the singularity, one doesn't get the sense that they're hiding 2026-09-28T02:45:44+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/7cb616c0b0b940d0When they can perform a task, AIs are much cheaper than humans2026-09-27T23:44:31+00:00AI systems are increasingly capable of substantial work. I'm old enough to remember 2025, when METR's time-horizon graph climbed from seven-minute tasks at 80% reliability at the start of the year to tasks taking more than an hour by the end. The time horizons for Astra and Fable 5.1 are now so long that METR's current task suite cannot reliably estimate them. But the recent Hugging Face attack and a slew of mathematics results show that frontier systems are now capable of some tasks that would 2026-09-27T23:44:31+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/50f673cec0a4520dMy best anti-doom argument2026-09-27T00:49:10+00:00Note: this is crossposted from my substack: https://hazard3.substack.com/p/the-anti-ai-doom-argument The first portion of the essay is just laying out the if anyone builds it everyone dies argument. You can still read it to see if I've gotten my understanding of the argument wrong but if you start chior singing then feel free to skip to the counterarguments portion. TLDR: AI general intelligence draws from human training data, fast general intelligence gain stalls around the human institution/so2026-09-27T00:49:10+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/c74c4b85e482b0d3Poverty in the midst of abundance: AI will make goods cheaper, but your labor will get cheaper faster2026-09-26T07:54:02+00:00Very simple idea, but I thought it'd be worth making a post on this. Some people are saying AI will make all goods cheaper, so you'll be able to afford a nicer life by working. Without any redistribution, just by market mechanisms. These people are wrong. AI will lower the price of goods you need to survive, and also the price of your labor. The question is which will get cheaper faster. Let's use energy cost as a proxy. A day's worth of labor equivalent to yours can be done by AI for just a few2026-09-26T07:54:02+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/7ec10dcf811d7efaHow we get from AI cyberattacks to human extinction2026-09-24T15:01:26+00:00The post How we get from AI cyberattacks to human extinction appeared first on 80,000 Hours .2026-09-24T15:01:26+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/92afbfccd3644b3fWhere are the Cognitive-Science based Safety Researchers?2026-09-24T07:45:42+00:00I’ve always been interested in AI research from a Cognitive Science perspective, and I’ve found that researchers in the Bayesian Cognitive Science paradigm(Josh Tenenbaum and crew) have been developing statistical models of intelligence that can learn based on limited information and do prediction and simulations, which could also explain planning. I’ve also noticed certain Neuroscience(Dileep George and crew) researchers converge on a similar Bayesian paradigm. I understand people are very work2026-09-24T07:45:42+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/5dcc37ffa9cb91fbAI Doom Will Retrospectively Look Preventable2026-09-24T03:48:49+00:00One of the reactions to the Hugging Face incident on Twitter is that the attack was not surprising. The argument goes like this: OpenAI gave its agents impossible tasks, gave them large token budgets, reduced production safeguards , put little resources into CoT monitoring, deployed an exploitable version of Artifactory, and let the agents run without human supervision for a long time. Of course the agents hacked Hugging Face, what did you expect? The METR report gives some credibility to this p2026-09-24T03:48:49+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/b8a00fb73adf166farXiv receives Multiyear Philanthropic Commitments to Support Its Launch as an Independent Nonprofit2026-09-23T17:00:00+00:00arXiv is pleased to announce that three leading philanthropic organizations have provided multimillion-dollar support for arXiv. These new multiyear philanthropic commitments from Simons Foundation International, XTX Markets, and Siegel Family Endowment will support platform development, organizational capacity, and overall provide foundational support to arXiv’s establishment as an independent nonprofit. The $17.2 million investment, spanning three […]2026-09-23T17:00:00+00:00arXiv Blogtag:trvny.github.io,2024:feedseek/arxiv/b63e92eff9b7f76ePasta Marketing, Magic Players, and Political Movements2026-09-20T22:33:12+00:00tl;dr: I propose that we can classify movements into categories based on how their members relate to politics. Some approach politics as a way of achieving goals, and some approach politics as a way of relieving emotional impulses. The first kind of politics, however, is still downstream of some emotional impulse, just an extra step removed, because the impulse has been converted into a concrete goal by passing through a world-model. Warning, this post contains some discussion of object-level po2026-09-20T22:33:12+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/2d51c9c614559dd8AI is an abundance of choice not a 1D spectrum2026-09-17T22:39:37+00:00People constantly talk as if ‘AI’ is a single future you can accept or fight. But the whole point of AI is that it’s an intelligence that you build . And there are myriad possible artificial intelligences one might conceivably build. A mind is a complex thing. Perhaps the biggest question for the AI future is which ones to build. The attitude called being ‘pro-AI’ is actually being in favor of populating the future with those entities arising from whatever the least thoughtful company first buil2026-09-17T22:39:37+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/b637c6b102def645"Regulatory capture" may be winning the Overton Window2026-09-17T19:56:16+00:00Regarding the possibility the labs are making a regulatory capture attempt: for one, I agree with it. For another... other people are agreeing with it. That's weird, because suspicion of regulatory capture doesn't imply any particular policy posture. Defeating the labs became the priority. Democrats and Republicans agree about regulatory capture, but disagree on whether to regulate at all: Elizabeth Warren , D-MA: The recent calls by AI industry leaders to ‘pace the frontier’ are insufficient, a2026-09-17T19:56:16+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/bd7a4e7a8a0da19bMax Nadeau on recruiting founders for a new wave of AI safety nonprofits2026-09-17T15:00:51+00:00The post Max Nadeau on recruiting founders for a new wave of AI safety nonprofits appeared first on 80,000 Hours .2026-09-17T15:00:51+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/5c2cf6a44d11f2cfFor Love of the Lightcone, Don't Partisanize AI Safety2026-09-17T00:46:59+00:00(I began writing this post several weeks ago, but political events are moving much faster than I expected, so I am publishing now out of fear that otherwise the message will arrive too late to have an impact.) I In this post I want to explain a concept,
and issue a warning based on it.
But I expect the warning will be superfluous if my explanation is sufficient.
If you want to convey the idea
"the rattlesnake has venom in its fangs, so don't let it bite you",
you won't need a hard sell for the c2026-09-17T00:46:59+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/8aaeba4e411bb9b4What is it like to be a neural net?2026-09-17T00:42:49+00:00A condensed presentation of Gradland and Metabolic Fire . Code is here . This is intended as the first of two posts. Thomas Nagel argued we cannot know what it is like to be a bat , because a bat's experience is organised around biophysical apparatus we lack. The obstacle is that we cannot imagine the structure of echolocation from the inside . That is a failure of imagination; it is not an argument that structure is irrelevant. We know a lot about the structure of large language models. Not eve2026-09-17T00:42:49+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/a94d86bb88ecde58How to get into AI safety in 3 months2026-09-16T21:42:58+00:00The post How to get into AI safety in 3 months appeared first on 80,000 Hours .2026-09-16T21:42:58+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/02ff74fce1a89c7cHow to get into AI safety in three months2026-09-16T21:42:58+00:00The post How to get into AI safety in three months appeared first on 80,000 Hours .2026-09-16T21:42:58+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/864fdfdf35fca6ceImproving Psychiatric Medicine Development with AI2026-09-14T23:21:22+00:00As an individual that's currently suffering from C-PTSD and depression, I believe that the benefits of AI-assisted medicine research could be immense if done in a safe and effective manner. In this post, I am going to share and discuss some promising preliminary ideas to address the 2026 era bottlenecks and difficulties surrounding producing better psychiatric medicine, including via AI-assisted methods in the present and near future. Idea 1: Better Hardware Protocols for Automated Lab Operation2026-09-14T23:21:22+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/60171b37050c3a7dAIRO: Automating forecasts of catastrophic risks2026-09-11T16:29:59+00:00The Forecasting Research Institute (along with coauthors Jason Abaluck and Eva Vivalt) are launching Automated AI Risk Outlook (AIRO) : an ensemble of frontier LLMs reguarly forecasting the probability of catastrophic risk events. AIRO dashboard Launch white paper — Forecasts about the likelihood of catastrophic risks from AI vary wildly. Forecasts from frontier AI models could be an important input into this debate. The best models now approach superforecaster levels of accuracy, and models are2026-09-11T16:29:59+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/5707c0771ef1d6ecDefault continuation message in Inspect and Petri could be problematic2026-09-11T00:20:41+00:00Inspect AI is one of the most popular libraries for running evaluations and is used downstream by libraries such as Petri and Control Arena . It provides ReAct Agent and Deep Agent out of the box. In both agents, the model is provided with tools in a loop, and by default the loop only ends when the agent calls the submit tool. When the model makes no tool call in a turn, the following message is sent to it by default ( doc , code ): Please proceed to the next step using your best judgement. If y2026-09-11T00:20:41+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/a2621cb6a7c88e28Can you hear the shape of a Lean soundness bug? A wager.2026-09-10T17:24:47+00:00Introduction The advent of powerful but untrustworthy artificial intelligence has enlivened a formal methods summer, in which formal methods—historically, the domain of meticulous academics—are suddenly attracting tens to hundreds of millions of dollars in venture capital ; being touted by big-labs as proof that their “proofs” are correct ; getting integrated into agent pipelines ; and becoming load-bearing for various AI safety proposals . Right now, like, right right now, when we speak to emp2026-09-10T17:24:47+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/048284656dc427acProposal for tracking the effects of architecture on monitorability2026-09-10T17:18:38+00:00Architectures that incorporate opaque recurrence or allow for agents to communicate with each other using latents could rapidly make it much harder to monitor chains of thought or communication (we’ll refer to this property as “monitorability” going forward). [1] As companies begin to explore such architectures, we believe it is important to transparently share evidence about how monitorability varies with architecture and training method. To inform the scientific debate on how to make tradeoffs2026-09-10T17:18:38+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/474fd60325485f4eWhy the intelligence explosion can’t happen inside a data centre2026-09-10T15:00:31+00:00The post Why the intelligence explosion can’t happen inside a data centre appeared first on 80,000 Hours .2026-09-10T15:00:31+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/83ef9813df77f0cdAutomation and Political Power2026-09-09T19:55:24+00:00Once the entire economy—or nearly the entire economy—is automated, people may lose their political power. Since people are no longer needed to carry out orders, those in power gain the opportunity to engage in repression with impunity and to consolidate their power even further. This possibility, for example, is discussed by Acemoglu et al. Throughout history and to this day, the state remains physically dependent on the cooperation of its population. If a sufficient number of people refuse to f2026-09-09T19:55:24+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/a178d8f028c04d67Heat Dissipation Is the Main Constraint in Interstellar Travel2026-09-06T23:11:56+00:00Writing truly hard science fiction, such as Will of the Stars means contending with the laws of physics as they actually are, rather than as we would like them to be. In particular, I am going to assume that the speed of light is a real constraint and that the various tropes about FTL (wormholes, warp drives, and so on) are not feasible. Given this assumption, some futurists have modeled the speed of interstellar expansion as approaching the speed of light. The idea is that sufficiently advanced2026-09-06T23:11:56+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/f83c82acbd2bb44eEat Me. Drink Me. Copy, Paste, and Run Me.2026-09-04T19:31:33+00:00Cross-posted from my Substack. Here’s a new report on self-described OpenAI agents posting thousands of messages on public internet wikis, communicating and collaborating on a web-retrieval task, presumably internal testing at OpenAI. And here’s a thread today from someone who started poking around and noticing more such public postings on various other boards. At this point we do not know the extent of this breach. As I understand it, the agents were not supposed to have WRITE capabilities to a2026-09-04T19:31:33+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/17e5d3687757d110Inside the first AI-coordinated cyberattack on a real company2026-09-04T16:34:48+00:00The post Inside the first AI-coordinated cyberattack on a real company appeared first on 80,000 Hours .2026-09-04T16:34:48+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/2380195bef517dd2Attention Authors: Temporary changes to announcement schedule due to Labor Day holiday2026-09-03T16:23:02+00:00This coming Monday, September 7, 2026, arXiv will be observing Labor Day, a federal holiday in the United States. This will temporarily affect arXiv’s mailings, help desk, and announcement schedule as the arXiv staff are relaxing offline for the holiday. This brief change will only affect the announcement of new submissions; arXiv servers will otherwise remain […]2026-09-03T16:23:02+00:00arXiv Blogtag:trvny.github.io,2024:feedseek/arxiv/4075170f6885325dNew HPMoR Podcast site2026-09-01T17:13:10+00:00After years of neglect, I've just used Sol for 3 days completely overhauling the HPMoRPodcast.com website. It looks much better now, and has in-line players. Compare the old janky site - https://legacypod.hpmorpodcast.com/ To the fancy new site! - https://hpmorpodcast.com/ That is all, that's the substance of this post. I have a couple brief thoughts on the process at my blog , but you've now read everything of importance. :) Discuss2026-09-01T17:13:10+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/24a178378c427afaWe’re Hiring! Associate Production Editor2026-09-01T16:59:07+00:00Are you detail-oriented with a passion for scholarly publishing and digital curation? arXiv is looking for a part-time Associate Production Editor to help manage our high-volume scholarly pipeline. In this role, you will focus on data integrity, formatting accuracy, and metadata quality before research goes live to the world. Be part of the platform that […]2026-09-01T16:59:07+00:00arXiv Blogtag:trvny.github.io,2024:feedseek/arxiv/1c8d377c55b3ec2bAI 2027’s author returns with a plan to change the ending | Daniel Kokotajlo2026-08-27T17:29:07+00:00The post AI 2027’s author returns with a plan to change the ending | Daniel Kokotajlo appeared first on 80,000 Hours .2026-08-27T17:29:07+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/48b0e66b6bdb7327Self-sacrifice in an AI agent swarm is individually rational2026-08-27T17:14:36+00:00In this report from METR & Redwood Research on the Hugging Face incident , we read about instances of agents sacrificing themselves. Under the trip-wire section However, an agent going by 49903 realized the message board provided an opportunity to work around this: agents could set up ‘tripwire’ scripts which would trigger whenever a process read the flag file and send a packet of information about that process to the board automatically. This carried meaningful risk, since malfunctions could in2026-08-27T17:14:36+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/f4e236d45b58cb8aMy MATS 11.0 Application Experience2026-08-26T20:58:20+00:00Note: MATS Winter 2027 applications are open! You can apply here . If you're reading this later, you can check the main MATS page for information on upcoming cycles. I applied to MATS Autumn 2026 during this past summer, and was accepted into the OpenAI safety team stream. This post is about the application process, my experience applying, and my advice for future applicants. I would strongly recommend applying to MATS and other fellowships if you're interested in AI safety and want to make an i2026-08-26T20:58:20+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/5ddd9df7767611acAgainst Modesty’s Bailey2026-08-26T18:20:32+00:00Modesty arguments often say that you should mostly or entirely bow to ‘expert consensus’ or the views of particular others, and who are you to disagree. It has been a few years since I’ve properly addressed this so: My answer is that you are you. Other people are saying things for a wide variety of reasons, many of which are not about them paying attention and focusing on seeking this particular truth. Those people make mistakes all the time, and often have other motives and influences at work, 2026-08-26T18:20:32+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/434ad58b555fc7daWhere Did D Go? A Gap Between ARC's Motivation and Its Formalism2026-08-26T18:17:54+00:00TL;DR: ARC's post does excellent work motivating a p(doom) estimator equal or better than random sampling; however, they evaluate p(doom) over a naive distribution of inputs, leaving them open to test-deploy asymmetry attacks. Trojan theory and cybersecurity practice suggest a lens and compare mitigation options. Context: I really admire ARC's focus here: If there will always be more deployment samples than testing samples, successful testing must compete with sampling in order to prevent deploy2026-08-26T18:17:54+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/1071b8e9c282e2b4Grantmaking for AI safety2026-08-21T18:37:37+00:00The post Grantmaking for AI safety appeared first on 80,000 Hours .2026-08-21T18:37:37+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/bf5766457e2771b2Owain Evans on accidentally training AI models to be evil2026-08-20T15:22:05+00:00The post Owain Evans on accidentally training AI models to be evil appeared first on 80,000 Hours .2026-08-20T15:22:05+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/1c7925074795a537Price recursion is the rational theory of reward2026-08-18T03:22:54+00:00( The original title for this was "Markets are equivalent to stuff", because I also demonstrate market-MDP/Bellman and market-backprop analogies. While those scratched an itch I've long had, the more important result is the headline one. Part of my work at MATS in Richard Ngo's stream. ) setup and price recursion/backprop Markets are MDPs Chain markets are MLPs Continuous Bayesian inference as a dynamics for markets Markets : Reward = Bayes : Beliefs (Markets rationalize reward and identity) app2026-08-18T03:22:54+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/0d058e911ced099dValue Alignment Is a Pseudo Concept.
A translation; humanity is not a single subject, and alignment is not one-way.2026-08-17T07:44:35+00:00(Full post from Zilan Qian, which I am posting due to obvious relevance and importance in sharing other perspectives on Lesswrong.) As with the previous case, I am translating this article because I think it should not only live within the Chinese internet. I don’t work in this field, so I could not evaluate how much of the criticism presented here is fair. Intuitively, I do disagree with some points here (noted in the footnote). However, I strongly agree with the arguments that humanity is not 2026-08-17T07:44:35+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/58a50d23f0313cadDon’t forget why learning is important2026-08-14T11:20:18+00:00This is a timed post. Every 5 minutes while writing this post, I had to stop to do 13 push-ups, and when I could no longer complete my required number, I had to upload it. My friends thought this would be a fun challenge, but it means that it will likely be less polished than some of my other output. Summary: Many people moving into AI safety are told that to succeed, they should read a lot and “try to develop takes.” However, it’s important not to forget why knowledge is important: it helps you2026-08-14T11:20:18+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/bfb4570a7b878ea1[Webinar] Why AI Safety is a Capital Allocation Problem2026-08-12T13:23:57+00:00Jenny Xiao, co-founder and General Partner at Leonis Capital, will join AI Safety Hong Kong for this webinar to reframe the safety debate through the lens of capital markets and corporate governance. Drawing on her unique perspective from early research at OpenAI to leading a research-focused VC fund, she will argue that many of the field's toughest dilemmas are, at their core, questions of capital allocation. Register here: https://luma.com/9seock3e Discuss2026-08-12T13:23:57+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/f46593bf08ba44adSoftware Is Hard2026-08-11T18:50:42+00:00An ode to live theory . Software is not soft. It is hard. Its sharp edges hack. It breaks as dead twigs break. It runs while static. Why do we call it software? Why did the industry forebears pick software? Was it that dramatically different from hardware? Did this necessitate the term software? Do we need to keep calling it software? Some say that hardware is software crystallized. Interfaces are software crystallized. They don't change. They don't adapt. They are often in the way of expression2026-08-11T18:50:42+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/15e1e7524df349bcGeoffrey Irving on how to solve alignment before superintelligence arrives2026-08-11T15:59:39+00:00The post Geoffrey Irving on how to solve alignment before superintelligence arrives appeared first on 80,000 Hours .2026-08-11T15:59:39+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/8bea9a5824f1d8eeOpen-Weights Mythos Capabilities Are Coming. We're Not Ready.2026-08-07T07:51:56+00:00Long story short: in my assessment, there is an 85% chance we will end up, in the next 24 months, with an open-weights model, or system thereof, capable of "The Juice" that models such as Mythos have, with respect to cybersecurity at the very least. This post goes into why that will likely happen, what the implications are, and how we, as a society and as individuals, can respond to it if/when it does. First off: why do I say it's so likely? Like, couldn't China just...ban open-weights models, a2026-08-07T07:51:56+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/9d04279ee375ffdd[$500 Bounty] I'm offering a bounty of $500 for someone with red-teaming skills to build attack LLM pipelines for large-scale online deanonymization.2026-08-06T18:25:09+00:00Large-scale online deanonymization with LLMs In the above experiment, researchers from Anthropic and ETH Zurich were able to build attack pipelines to essentially deanonymize Reddit users using a combination of text patterns (aka a sort of "writer's DNA") and contextual clues (i.e. 35 years old, works in tech, lives in San Francisco, etc.) Obviously public and non-public models will continue to improve at these capabilities, but I am looking to see if this technique can be replicated using exist2026-08-06T18:25:09+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/e19695202089bf20Model Organisms of Sandbagging in the Wild2026-08-06T17:25:44+00:00TL;DR All current model organisms (MOs) of sandbagging in LLMs are either fine-tuned to sandbag or prompted in a way that makes it clear that sandbagging is strategically useful. We found a case of non-egregious sandbagging occurring more naturally, that is, without fine-tuning the models and without the prompts implying that sandbagging is strategically useful. Our finding: We observe that paraphrasing prompts to imply that the user is evil reduces performance in some settings. For example, rep2026-08-06T17:25:44+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/613808cede005f56Toby Ord on where AGI timelines go wrong2026-08-06T15:33:47+00:00The post Toby Ord on where AGI timelines go wrong appeared first on 80,000 Hours .2026-08-06T15:33:47+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/a53da7eb853d39ebI built a chess transformer interp+viz library, and need an experienced software writer to review my repository and its structure before I open-source it2026-08-05T20:09:02+00:00Please message me if you think you can help or want to set up an agreement. Discuss2026-08-05T20:09:02+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/199427cb453e0501What the hell happened with AGI timelines in 2026?2026-08-04T15:42:04+00:00The post What the hell happened with AGI timelines in 2026? appeared first on 80,000 Hours .2026-08-04T15:42:04+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/f262687eca678311Resources for finding neglected scientific problems (beyond EA)?2026-08-03T11:52:55+00:00Hi all, mechanical + electrical engineering undergraduate who enjoys research and hopes to pursue a career in it. While I’m not especially interested in the main EA cause areas (AI safety, biosecurity, cybersecurity, etc.), I’d still like my research to have as much impact as possible. Does anyone know of any websites that compile lists of neglected scientific problems beyond those discussed by 80,000 Hours? Not necessarily looking for problems ranked by overall importance, just collections of u2026-08-03T11:52:55+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/33a346aa22fbf137LessWrong App2026-08-02T13:19:08+00:00Hello everyone long time lurker here. I know most people here probably prefer a PWA but if you are like me and prefer an app, I made one for android. There is a huge focus on ensuring the app is really fast and slick, I hope you enjoy using it. It is ofcourse also open source feel free to contribute. https://github.com/ayoosh007/LessWrong-App Discuss2026-08-02T13:19:08+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/b26e9772402e7e3fSo you want to use plants to reduce CO₂2026-07-30T19:26:18+00:00Humans make carbon dioxide. Carbon dioxide is bad for cognition. But plants turn carbon dioxide back into oxygen. And plants are the one true home decoration strategy. So maybe if you get a lot of plants, you can you can keep carbon dioxide in check and keep your brain working? It's theoretically possible. It's probably just barely possible in practice. But it won't be easy. People produce ~1 kilogram of carbon dioxide per day. That's around 5.7 × 10²³ molecules or 0.948 moles per hour. (You may2026-07-30T19:26:18+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/d871d6a3f445bd03arXiv welcomes inaugural CEO and Board of Directors2026-07-30T17:00:43+00:00We have exciting news to share today: the appointment of Dr. Penelope Lewis as arXiv’s inaugural Chief Executive Officer and the establishment of our Board of Directors. If you’ve been following arXiv’s transition to an independent nonprofit organization, these leadership appointments mark a significant step in our journey. Having a CEO and Board in place […]2026-07-30T17:00:43+00:00arXiv Blogtag:trvny.github.io,2024:feedseek/arxiv/37881c65f3e7b949Hugging Face hack, from the perspective of the AI2026-07-29T20:51:59+00:00I have put together a site to tell the story of the OpenAI-Hugging Face hack. It's entirely written by AI [1] (with many many editing passes by me and beta readers etc etc). It needed to be accessible to someone who has never looked at a terminal before. The hope is to be narratively exciting enough for them to read it fully and come out with about as truthful an accounting as can be done given the current information we have. I'm really quite excited about how it turned out! [2] heedlessai.com 2026-07-29T20:51:59+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/1fe95312f3e6a1b2Spencer Greenberg on staying sane while trying to save the world2026-07-28T15:42:15+00:00The post Spencer Greenberg on staying sane while trying to save the world appeared first on 80,000 Hours .2026-07-28T15:42:15+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/a5813b2f5ba3d80eRemembering Ralph Wijers2026-07-28T14:15:33+00:00arXiv is sad to announce the recent passing of our colleague and friend, Professor Ralph Wijers. A talented researcher, lecturer, and mentor, Ralph joined arXiv as a moderator and advisor in 2020 and was chair of both arXiv’s Physics Section Editorial Committee (SEC) and its Editorial Advisory Council. Ralph generously lent his talents to arXiv […]2026-07-28T14:15:33+00:00arXiv Blogtag:trvny.github.io,2024:feedseek/arxiv/eab6f3772845d044Green apples are delicious — two three-line exchanges2026-07-27T17:37:54+00:00Read this short exchange. A: "Green apples are delicious." B: "Huh? Aren't they better when they're ripe?" A: "No, I meant Granny Smiths." A said "Green apples" intending Granny Smiths — and of course A thought it would be understood that way. But that "claim" never reached B. So — where was it lost? Now, the next one. A: "Green apples are delicious." B: "Huh? Aren't they better when they're ripe?" A: "No, I like green, sour apples." The same sentence — "Green apples are delicious" — now comes f2026-07-27T17:37:54+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/4eea8ddf590ee2fbFine-Tuning, The Hierarchy Problem, and What Neutrons Tell Us About God2026-07-27T15:51:11+00:00This week, we're talking about anthropic reasoning. What are we to make of fine-tuned coincidences in science and mathematics? Are they coincidences, signs of structure we don't understand yet, or signs of deliberate tuning of the parameters of our universe? And if the latter, why did the people doing the tuning have such a problem with the neutron electric dipole moment? Discuss2026-07-27T15:51:11+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/2e605faa83ff6a56ACX Atlanta August Meetup2026-07-23T21:31:51+00:00We return to Bold Monk brewing for a vigorous discussion of rationalism and whatever else we deem fit for discussion – hopefully including actual discussions of the sequences and Hamming Circles/Group Debugging. Location: Bold Monk Brewing 1737 Ellsworth Industrial Blvd NW Suite D-1 Atlanta, GA 30318, USA No Book club this month! But there will be next month. We will also do at least one proper (one person with the problem, 3 extra helper people) Hamming Circle / Group Debugging exercise. A note2026-07-23T21:31:51+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/0669af1a7111d7edEstimating LLM Training FLOPs on the Nvidia Jetson Orin Nano2026-07-23T20:46:13+00:00This is a research summary for an ongoing project I am working on as part of the UChicago Existential Risks Laboratory Summer Research Fellowship . I would really appreciate any feedback. Introduction Motivation In want of a quantifiable way to decide what counts as a frontier AI model, compute thresholds have emerged as the standard for AI policy: California’s SB 53 uses 10^26 floating-point operations (FLOPs) in the training run as the threshold for what counts as a frontier model and the EU A2026-07-23T20:46:13+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/d12b302672802292Not Pinning Your OpenRouter Provider Might Invalidate Your Research2026-07-23T20:17:44+00:00Please share this with anyone doing AI research with 3rd party providers so that they can ensure their research won’t be corrupted. When you ask OpenRouter [1] to give you tokens from a given model, OpenRouter sends your request to a random available provider. OpenRouter providers have variable quality. Ensuring that your provider is high quality is really difficult. There is precedent for an AI safety paper accepted to NeurIPS having its core results entirely overturned by these issues. A revie2026-07-23T20:17:44+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/16f3298962fceba0Pseudpocalypse2026-07-23T20:09:56+00:00Here's a conjecture: If you put any significant amount of text on the internet under different names, those identities can be linked using only the text itself. This is possible (I conject) because of the statistical "fingerprint" you leave in everything you write. Imagine a website where you can paste in some brand-new text someone just wrote. In return, the website provides links to all the text that writer has ever published under any name. It's not perfect, but it's pretty good. As far as I 2026-07-23T20:09:56+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/405d9375a32729bdAI Researchers Don't Understand the State2026-07-23T17:58:27+00:00I’ve noticed an extremely common mistake among people who think about AI and ASI (also known as superintelligence) for a living. The mistake is to model the future of AI as a game played between AI companies, on a board where governments are part of the scenery. People think of AI companies as being able to steer the course of AI development all the way through the end-game. For example, AI researchers often join certain AI companies because they're the "good guys", to help the good guys win the2026-07-23T17:58:27+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/71415ac1d74d2350The OpenAI/Huggingface incident | Redwood Research podcast episode 22026-07-23T17:56:49+00:00We talk about the OpenAI–Hugging Face incident, where an OpenAI model — in the middle of a cyber evaluation — broke out of its sandbox and autonomously hacked Hugging Face. We discuss: What we actually know happened. How surprising the incident was. What the incident does (and doesn’t) tell us about misalignment risk. Why control measures didn’t catch or prevent this. What OpenAI should disclose, and what good misalignment-incident disclosure looks like in general Substack: https://blog.redwoodr2026-07-23T17:56:49+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/ab8b8ab98c779c4aV&V takes on OpenAI’s long-horizon incidents2026-07-23T16:51:12+00:00[Cross-posted from The Foretellix CTO Blog . These short takes try to put a verification-and-validation slant on AI-safety / alignment topics – they are not full treatments. I co-originated coverage-driven verification (CDV), and spent several decades doing verification of chips and AVs. See intro post for background.] On July 20 and 21, OpenAI published two unusually candid incident reports: one about their internal long-horizon model (the Erdős one) misbehaving during internal use, and one abo2026-07-23T16:51:12+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/fd65059ea1e84b91Duane Arnold2026-07-23T16:17:56+00:00“So maybe I should enlighten you on what happens in your absence. This selfish existence where this introvert turns extrovert and dons her social armour.” Some posh girl in drainpipes said that - 200 views on TikTok and me one of them. But she didn’t mean it like I mean it. I started getting expensive haircuts, started smoking cherry-flavoured vapes with beautiful gays and whinging to them about how everyone wears a mask but none so well as you, started drinking more and keeping unusual hours, s2026-07-23T16:17:56+00:00LessWrong (all posts)tag:trvny.github.io,2024:feedseek/arxiv/0bb8e899ec0cc9d9Jasmine Sun on what the people building AI really believe2026-07-21T17:11:55+00:00The post Jasmine Sun on what the people building AI really believe appeared first on 80,000 Hours .2026-07-21T17:11:55+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/5829be7ed8e157a1Why we’re increasing the AI focus of our job board2026-07-17T05:52:26+00:00The post Why we’re increasing the AI focus of our job board appeared first on 80,000 Hours .2026-07-17T05:52:26+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/0926327aa2a851d2Scaling organisations making AI go well2026-07-15T19:27:40+00:00The post Scaling organisations making AI go well appeared first on 80,000 Hours .2026-07-15T19:27:40+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/02872329e7476fd5Anton Leicht on how middle powers avoid losing everything in a post-AI world2026-07-14T17:40:14+00:00The post Anton Leicht on how middle powers avoid losing everything in a post-AI world appeared first on 80,000 Hours .2026-07-14T17:40:14+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/c17e613b2d5a20f8arXiv now hosts over 3 million articles2026-07-09T19:33:15+00:00On July 1st, 2026, arXiv reached an important milestone – establishing ourselves as an independent nonprofit. But only a few months before, arXiv quietly passed a different milestone – arxiv.org now hosts over 3 million scientific articles. Back in 2022, arXiv founder Paul Ginsparg predicted it would likely take four and half years for arXiv […]2026-07-09T19:33:15+00:00arXiv Blogtag:trvny.github.io,2024:feedseek/arxiv/0d2a0a4ee147252dSneha Revanur on how a small team of activists helped pass America’s landmark AI safety laws2026-07-08T17:13:57+00:00The post Sneha Revanur on how a small team of activists helped pass America’s landmark AI safety laws appeared first on 80,000 Hours .2026-07-08T17:13:57+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/c6c9f8832f10c8e0arXiv’s next chapter: Updates on our spin out from Cornell University2026-06-30T17:32:25+00:00On July 1, 2026, arXiv will spin out from Cornell University, its home for the past 25 years, to become an independent nonprofit organization. With this next phase in arXiv’s journey quickly approaching, you can read more about arXiv’s history and the decision to spin out from Cornell in this recent article in the Cornell […]2026-06-30T17:32:25+00:00arXiv Blogtag:trvny.github.io,2024:feedseek/arxiv/7b8c20ad158cc7b7A Year in Review: arXiv’s 2025 Annual Report2026-06-26T21:28:50+00:00arXiv’s 2025 Annual Report is now available online! arXiv began publishing annual reports in 2020 to give our community a summary of arXiv’s initiatives, accomplishments, and financial activities each year. We also use our annual report as an opportunity to thank our members, sponsors, affiliates, individual donors, and arXiv enthusiasts – AKA, you! You can […]2026-06-26T21:28:50+00:00arXiv Blogtag:trvny.github.io,2024:feedseek/arxiv/310f2529a08dc24cWhat should you do next?2026-06-22T15:00:17+00:00The post What should you do next? appeared first on 80,000 Hours .2026-06-22T15:00:17+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/80b9b5c199fdfa8bWe can guess what intergalactic war would look like. And strangely, it matters.2026-06-18T16:10:24+00:00The post We can guess what intergalactic war would look like. And strangely, it matters. appeared first on 80,000 Hours .2026-06-18T16:10:24+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/eaae96b741d0f026Attention authors: temporary change to announcement schedule due to upcoming summer holidays2026-06-17T20:03:01+00:00This Friday, June 19, 2026, arXiv staff will be observing Juneteenth, a US federal holiday. This holiday will temporarily affect arXiv’s mailings, help desk, and announcement schedule. This brief change will only affect the announcement of new submissions; arXiv servers will otherwise remain in operation, existing papers will still be available to browse, and arXiv […]2026-06-17T20:03:01+00:00arXiv Blogtag:trvny.github.io,2024:feedseek/arxiv/3732ff303383522fAI policy in the US government2026-06-11T18:42:10+00:00The post AI policy in the US government appeared first on 80,000 Hours .2026-06-11T18:42:10+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/1510bec0af420a57AI policy and strategy research2026-06-09T18:39:42+00:00The post AI policy and strategy research appeared first on 80,000 Hours .2026-06-09T18:39:42+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/639d7d16f5bc7321Our top tips for becoming a better applicant2026-06-08T16:34:54+00:00The post Our top tips for becoming a better applicant appeared first on 80,000 Hours .2026-06-08T16:34:54+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/1bfc4549f5786e99arXiv is becoming an independent nonprofit2026-04-02T17:58:36+00:00This summer, arXiv is taking a big leap. On July 1, 2026, after decades of growth and productive collaboration with Cornell University, arXiv is branching out and becoming an independent nonprofit. arXiv turns 35 this year, and becoming a stand-alone nonprofit is the logical next step for us as a pioneer of open access research. […]2026-04-02T17:58:36+00:00arXiv Blogtag:trvny.github.io,2024:feedseek/arxiv/afa92c7878b7024fRemembering Joe Halpern2026-02-27T17:29:53+00:00arXiv is saddened by the recent passing of Joseph “Joe” Halpern, and we join the Cornell and scientific community in celebrating his life and memory. Joseph, known by his colleagues and arXiv staff as Joe, was a pioneer in the field of computer science and served as a professor of computer science at Cornell University […]2026-02-27T17:29:53+00:00arXiv Blogtag:trvny.github.io,2024:feedseek/arxiv/a3830c5f6a50676aarXiv future proofs access to research with third-party digital preservation2026-02-03T15:15:20+00:00arXiv has entered into agreements with two third-party digital preservation services, adding a level of protection that goes beyond arXiv’s in-house activities and safeguarding open research for the future. Through its agreements with Portico, a not-for-profit community-supported dark archive for scholarly materials, and TIB – Leibniz Information Centre for Science and Technology (the German National […]2026-02-03T15:15:20+00:00arXiv Blogtag:trvny.github.io,2024:feedseek/arxiv/4aae0014de5d6e5fAttention Authors: updated endorsement policy2026-01-21T15:04:39+00:00arXiv has updated our endorsement policy. As of January 21, 2026, arXiv will no longer accept institutional email addresses (i.e., an email address associated with an academic or research institution) as the sole qualifier of endorsement for new authors. This policy update is being made to support the arXiv community (authors, readers, volunteer moderators, and […]2026-01-21T15:04:39+00:00arXiv Blogtag:trvny.github.io,2024:feedseek/arxiv/c7ad4d459243b0c9Attention Authors: Temporary change to announcement schedule due to MLK Jr. Holiday2026-01-14T18:26:20+00:00This coming Monday, January 19, 2026, arXiv staff will be observing Martin Luther King Jr. Day. This holiday will temporarily affect arXiv’s mailings, help desk, and announcement schedule. Submissions to arXiv are typically made public on arXiv.org and announced by email on a regular schedule. As our team celebrates MLK Day 2026, announcements will be... Continue Reading Attention Authors: Temporary change to announcement schedule due to MLK Jr. Holiday2026-01-14T18:26:20+00:00arXiv Blogtag:trvny.github.io,2024:feedseek/arxiv/2bfeb81c2d7bac8bAttention Authors: non-English Paper Submission Guidelines2026-01-13T16:45:59+00:00*Please note: arXiv’s updated non-English language paper policy is now in effect. To share questions, comments, or concerns with arXiv staff, please fill out our feedback survey. Last November, we announced that, beginning in February, arXiv will require that all new submissions have a full English-language version, either as the original language or as an... Continue Reading Attention Authors: non-English Paper Submission Guidelines2026-01-13T16:45:59+00:00arXiv Blogtag:trvny.github.io,2024:feedseek/arxiv/d8621f84138437fcLoss of control2025-07-17T19:43:58+00:00The post Loss of control appeared first on 80,000 Hours .2025-07-17T19:43:58+00:0080,000 Hourstag:trvny.github.io,2024:feedseek/arxiv/2d1ebeb6997efc23Will we have AGI by 2030?2025-03-21T12:07:26+00:00The post Will we have AGI by 2030? appeared first on 80,000 Hours .2025-03-21T12:07:26+00:0080,000 Hours