Search: “safety”

arXiv:2605.15514v1 Announce Type: cross Abstract: We identify intrinsic limitations of Rotary Positional Embeddings (RoPE) in Transformer-based long-context language models. Our theoretical analysis a

arXiv:2601.06366v2 Announce Type: replace-cross Abstract: Large Language Models (LLMs) are transforming enterprise workflows but introduce security and ethics challenges when employees inadvertently s

arXiv:2605.16024v1 Announce Type: new Abstract: Desktop GUI agents operate under partial observability: visually similar screens can correspond to different underlying workflow states, so locally plau

arXiv:2605.16165v1 Announce Type: cross Abstract: Autoregressive next-token training offers a unified formulation for image generation and text understanding, but it also creates strong modality compe

arXiv:2511.18719v4 Announce Type: replace Abstract: Reinforcement learning (RL) has become a powerful tool for post-training visual generative models, with Group Relative Policy Optimization (GRPO) in

arXiv:2511.17426v2 Announce Type: replace-cross Abstract: Self-supervised learning (SSL) has recently advanced through non-contrastive methods that couple an invariance term with variance, covariance,

arXiv:2605.15720v1 Announce Type: new Abstract: Medical referring image segmentation (MRIS) requires pixel-level masks aligned with textual descriptions of anatomical locations, making annotation cost

seriou question: how do you handle an intellectual doppelganger who has systematically started adopting every position you have argued for for 30 years while presenting each idea as if it were his own

arXiv:2605.16103v1 Announce Type: new Abstract: This paper develops a sign-separated finite-time error analysis for constant step-size Q-learning. Starting from the switching-system representation, th

arXiv:2604.02268v2 Announce Type: replace Abstract: Agent skills, structured packages of procedural knowledge and executable resources that agents dynamically load at inference time, have become a rel

arXiv:2605.15536v1 Announce Type: cross Abstract: Previous imitation learning policies predict future actions at every control step, whether in smooth motion phases or precise, contact-rich operation

arXiv:2605.16137v1 Announce Type: new Abstract: Generating simulation-ready tabletop scenes from task instructions is an intriguing and promising research direction in the field of Embodied AI. Howeve

arXiv:2605.15528v1 Announce Type: new Abstract: Autonomous underwater vehicle (AUV) swarms are emerging as intelligent underwater networks, where each node must sense, communicate, process local data,

arXiv:2505.15692v5 Announce Type: replace Abstract: Reinforcement learning (RL) has emerged as an effective paradigm for enhancing model reasoning. However, existing RL methods like GRPO typically rel

arXiv:2605.15517v1 Announce Type: new Abstract: We present a method for training reference-guided, perceptive reinforcement learning locomotion policies for humanoid robots in which reference trajecto

the AI trial of the century ended with a procedural whimper rather a bang; the jury agreed that Musk was too late but never weighed in on the questions of whether OpenAI did was legitimate. and so we

The pure LLM debate - which I had for many years, here and elsewhere - is indeed no longer relevant. Why? Because I won; nobody uses pure LLMs anymore. Nowadays all deployed objects are neurosymbolic,

The U.S. faces a chaotic #AI regulatory patchwork with over 1,200 state bills introduced in 2025 and no unified federal framework, researchers @JeffSonnenfeld and Stephen Henriques of @YaleSOM and @Ga

arXiv:2605.15611v1 Announce Type: new Abstract: Root cause analysis (RCA) in microservices is challenging due to (i) noisy and heterogeneous multimodal observability (metrics, logs, traces), (ii) casc

arXiv:2605.15649v1 Announce Type: new Abstract: Developing effective surrogates (performance predictors) for Neural Architecture Search (NAS) typically requires expensive fine-tuning or the engineerin

I can't verify this claim without access to the actual post and current information. The headline uses inflammatory language ('shook down') that suggests opinion rather than neutral reporting. To crea

arXiv:2605.15388v1 Announce Type: new Abstract: Stochastic estimators are fundamental to large-scale optimization, where population quantities must be inferred from noisy oracle observations. Although

arXiv:2505.11708v3 Announce Type: replace-cross Abstract: Reinforcement Learning (RL) agents are increasingly used to simulate sophisticated cyberattacks, but their decision-making processes remain op

arXiv:2605.15228v1 Announce Type: new Abstract: Modern cloud and enterprise systems rely on identity-centric authorization, assuming that callers possessing valid credentials are safe to execute comma

arXiv:2605.15458v1 Announce Type: new Abstract: Video diffusion models have made rapid progress in perceptual realism and temporal coherence, but they remain primarily optimized for plausible generati

arXiv:2605.15604v1 Announce Type: cross Abstract: Modern language models often need to optimize a primary accuracy objective while also accommodating secondary behavioral preferences, such as verbosit

arXiv:2512.00778v2 Announce Type: replace Abstract: Preference optimization (PO) is indispensable for large language models (LLMs), with methods such as direct preference optimization (DPO) and proxim

arXiv:2605.15959v1 Announce Type: cross Abstract: Physics-informed neural networks (PINNs) are powerful surrogates for differential equations but are notoriously difficult to train due to spectral bia

arXiv:2510.06062v2 Announce Type: replace Abstract: Reinforcement learning (RL) has shown great promise in large language models (LLMs) post-training, which typically rely on token-level clipping to m

arXiv:2605.15557v1 Announce Type: new Abstract: Continuous diffusion and flow models are attractive for non-autoregressive text generation because they can update all positions in parallel. A major di

You know how I said yesterday that my feed is littered with people just making stuff up? We didn’t have 71% of Americans opposed to cell phones. Or ipods. Or laptops. We *do* have 71% of Americans opp

Gary Marcus draws a historical parallel between Bernie Madoff's claimed 16.3% annual returns (which proved to be a Ponzi scheme) and Sam Altman's promised 17.5% returns, warning that similarly implaus

🚨Breaking new study: memory in LLM agents still can’t really be trusted, even after over trillion dollars has gone into the development of the field. Excited to share our new paper: “Useful Memories B

GDS weighs in on the NHS's decision to retreat from Open Source Terence Eden continues his coverage of the NHS' poorly considered decision to close down access to their open source repositories in res

profound comment by Daniel Eth: AI (long term) will probably change things a lot — but in ways that nobody can really predict. Your kids will not work 3.5 days a week or live to 100. They might work m

the whole point of databases is to keep reliable records of values that change over time. in general, they can be trusted to be stable over time. the whole point of LLMs is to raise money. in general

This is a nice article (not sure how I stumbled upon it a month later) I directionally agree with it in that: ✅ I have a massive bias for slope, grit, and scrappiness in candidates vs. pure experience

trillions. and i did try to warn them. repeatedly. for many years. 🤷‍♂️ @GaryMarcus everyone pretended reliability was a detail theyd fix later now we find out billions later it was always the core pr

Trump just got exposed for running the biggest insider trading operation in American history. Nancy Pelosi traded 5 million in stocks and Congress lost its mind. Trump literally executed 750 MILLION w

Dear @geoffreyhinton, You have to stop lying about me. First was the apparently faked quote on your web page that you couldn’t provide a source for (literally the only source I found was on your own w

Gary Marcus critiques AI companies for training large language models on copyrighted books and other proprietary content without permission or compensation to authors. He argues that these companies e

holy shit lmao @Gavriel_Cohen he's seriously using this thing for conducting the foreign policy/parliamentary affairs of singapore - and sharing his stack on how he is hacking around WhatsApp and doin

Gary Marcus discusses parallels between current AI development and societal trends with absurdist themes reminiscent of Kurt Vonnegut's work, suggesting that real-world outcomes increasingly resemble

Gary Marcus expresses optimism about the potential of LLM-powered AI agents in a post on X (formerly Twitter). The post advocates for confidence in these AI systems, though without access to the full

Cointelegraph: OpenAI partners with Malta's AI for All initiative to give citizens a free year of ChatGPT Plus if they complete a University of Malta AI literacy course — Cointelegraph is committed to

pleased to be in this and encourage everyone to read: Awesome new theme issue of Philosophical Transactions: ‘World models in natural and artificial intelligence’ Thank you @adamsafron! https://royals