Search: “safety”

GuideWalk: Learning Unified Autonomous Navigation and Locomotion for Humanoid Robots across Versatile Terrains

arXiv:2606.10449v1 Announce Type: new Abstract: Humanoid robots have achieved strong locomotion capabilities, but reliable navigation on versatile terrains remains challenging because obstacle avoidan

safetyarxiv-cs-ro

Safety

Hand-centric Human-to-Robot Trajectory Transfer from Video Demonstrations via Open-World Contact Localization

DGX agent

arXiv:2606.10743v1 Announce Type: new Abstract: Learning from human video demonstrations remains challenging due to noisy hand-object interactions, unseen objects with partial observation, and cross-e

safetyarxiv-cs-ro

10 Jun 2026

Hidden Consensus:Preference-Validity Compression in Human Feedback

arXiv:2606.10569v1 Announce Type: cross Abstract: Standard RLHF pipelines often reduce heterogeneous human judgments into a single scalar reward target. We argue that this reduction can mis-measure al

Hierarchical Policies from Verbal and Egocentric Human Signals for Natural Human-Robot Interaction

arXiv:2606.10276v1 Announce Type: cross Abstract: For natural human-robot interaction, a robot must understand human intent expressed not only through language but also through nonverbal signals such

HiGR: Industrial-Scale Hierarchical Generative Slate Recommendation Framework in Tencent

arXiv:2512.24787v3 Announce Type: replace-cross Abstract: Slate recommendation, which presents users with a ranked item list in a single display, is ubiquitous across mainstream online platforms. Whil

I was first to call the death of tokenmaxxing (ahead of @zerohedge), among the first (in 2023) to warn that the economics of Generative AI w…

I was first to call the death of tokenmaxxing (ahead of @zerohedge), among the first (in 2023) to warn that the economics of Generative AI was problematic, and first (also 2023) to warn that OpenAI co

IDEAL: In-DEpth ALignment Makes A Discrete Representation AutoEncoder

arXiv:2606.11096v1 Announce Type: new Abstract: Built on pretrained vision foundation models (VFMs), representation autoencoders (RAEs) have recently emerged as a promising approach for constructing s

If the new German court decision — to hold LLM companies liable for the false claims their LLMs made — become universal, effectively sidelin…

If the new German court decision — to hold LLM companies liable for the false claims their LLMs made — become universal, effectively sidelining generative AI until better techniques came along, would

Improving Adversarial Transferability on Vision-Language Pre-training Models via Surrogate-Specific Bias Correction

arXiv:2606.10571v1 Announce Type: cross Abstract: Adversarial examples reveal vulnerabilities in Vision-Language Pre-training (VLP) models and provide insights for improving robustness. A key property

Improving Text-Instance Alignment Of Foreground Conditioned Out-Painting Via Customized Concept Embedding

arXiv:2606.10892v1 Announce Type: cross Abstract: To showcase products, merchants often incur substantial costs creating high-quality display images. Foreground Conditioned Outpainting (FCO) meets thi

In addition to transparency, I now believe frontier models should face mandatory third-party testing for cyber, bio, and autonomy risks—with…

In addition to transparency, I now believe frontier models should face mandatory third-party testing for cyber, bio, and autonomy risks—with the power to block or revoke deployment of models that pose

safetydario-amodei--x

Informed Asymmetric Actor-Critic: Leveraging Privileged Signals Beyond Full-State Access

arXiv:2509.26000v3 Announce Type: replace Abstract: Asymmetric reinforcement learning leverages privileged information available during training to improve learning under partial observability. Existi

Inverse Probability Weighting and Age-of-Information Aggregation for Decentralized Federated Learning under Partial Reception

arXiv:2606.10774v1 Announce Type: new Abstract: Decentralized Federated Learning (DFL) over lossy wireless networks faces two key challenges: selection bias, where updates from poor-quality links are

Is Fairness Truly Fair? Towards Reliable Lipschitz Fairness in Multi-Task Learning via Fixed-exorpdfstring{elta}{delta} Alignment

arXiv:2606.10632v1 Announce Type: cross Abstract: Lipschitz-style individual fairness formalizes the idea that semantically similar examples should receive similar predictions, but its evaluation in m

It is not literally impossible to pause or slow down. It is a choice. But if the “frontier labs” continue to refuse to do, I strongly believ…

It is not literally impossible to pause or slow down. It is a choice. But if the “frontier labs” continue to refuse to do, I strongly believe it will eventually lead to tremendous harm. (Essentially e

It Takes One to Bias Them All: Breaking Bad with One-Shot GRPO

arXiv:2606.10931v1 Announce Type: new Abstract: Warning: This paper contains several toxic and offensive statements. Modern large language models (LLMs) are typically aligned through large-scale post-

LAFP: Preserving Latent Action Structure in Latent Policy Learning via Flow Matching

arXiv:2606.10517v1 Announce Type: new Abstract: Learning high-quality latent actions from large-scale unlabeled videos, coupled with limited real-world interaction data for training an action decoder,

Lightweight Latent Reasoning for Narrative Tasks

arXiv:2512.02240v2 Announce Type: replace Abstract: Large language models (LLMs) tackle complex tasks by generating long chains of thought or 'reasoning traces' that act as latent variables in the gen

Lip Forcing: Few-Step Autoregressive Diffusion for Real-time Lip Synchronization

arXiv:2606.11180v1 Announce Type: new Abstract: Diffusion-based lip synchronization models achieve strong visual quality and audio-visual alignment, but full-sequence bidirectional attention and many

LLM-Aided Joint Secrecy Precoding and Trajectory for RSMA-Based Heterogeneous UAV Networks

arXiv:2507.17188v2 Announce Type: replace-cross Abstract: This paper investigates secure communications in rate-splitting multiple access (RSMA) enabled heterogeneous UAV networks, where multiple UAVs

Locomotion analysis of a quadruped interacting with the lunar granular surface

arXiv:2606.10273v1 Announce Type: new Abstract: Deploying legged robots in extra-terrestrial environments includes many challenges due to complex terrain interactions, energy, and thermal constraints.

safetyarxiv-cs-ro

ManiSplat: Manipulation Trajectory Synthesis from Monocular Video via Decoupled 3D Gaussian Splatting

arXiv:2606.10645v1 Announce Type: new Abstract: Reconstructing dynamic and interactive 3D scenes from real-world observations remains a fundamental challenge in computer vision and robotics. While rec

Many of these policy ideas have common-sense appeal across the political spectrum, and the sooner we act on them, the sooner everyone shares…

Dario Amodei discusses policy proposals that have bipartisan appeal and advocates for prompt action to implement them for broad societal benefit. The post suggests that certain policy ideas transcend

safetydario-amodei--x

Masa can hype AI all he likes, but apparently he can’t even get a loan on his OpenAI shares. 🤔 But I actually half agree—and it might surpr…

Masa can hype AI all he likes, but apparently he can’t even get a loan on his OpenAI shares. 🤔 But I actually half agree—and it might surprise some of my critics to hear this, especially those who hav

Mean Flow Distillation: Robust and Stable Distillation for Flow Matching Models

arXiv:2606.11155v1 Announce Type: new Abstract: Flow Matching models have demonstrated strong performance across a wide range of generative tasks. However, their reliance on ODE-based iterative sampli

Measuring Human Value Expression in Social Media Texts: Calibrated LLM Annotation and Encoder Transfer

arXiv:2606.11018v1 Announce Type: new Abstract: Measuring subjective constructs in naturally occurring social media text requires annotation procedures that are theoretically grounded, empirically val

MedFeat: Model-Aware and Explainability-Driven Feature Engineering with LLMs for Clinical Tabular Prediction

arXiv:2603.02221v2 Announce Type: replace-cross Abstract: In clinical tabular prediction, classical machine learning models with feature engineering often outperform neural methods. LLMs are increasin

MIND-V: Hierarchical World Model for Long-Horizon Robotic Manipulation with RL-based Physical Alignment

arXiv:2512.06628v3 Announce Type: replace-cross Abstract: Scalable embodied intelligence is constrained by the scarcity of diverse, long-horizon robotic manipulation data. Existing video world models

Mitigating Bias in Low-SNR Financial Reinforcement Learning via Quantum Representations

arXiv:2606.10448v1 Announce Type: cross Abstract: The financial market is a typical low signal-to-noise ratio (SNR) setting, which often destabilizes off-policy maximum-entropy methods like Soft Actor

Mitigating Manifold Departure: Uncertainty-Aware Subspace Rectification for Trustworthy MLLM Decoding

arXiv:2606.09859v1 Announce Type: cross Abstract: MLLMs frequently hallucinate objects inconsistent with visual inputs. This issue is typically attributed to the over-reliance on language priors, whic

MMD Guidance: Training-Free Distribution Adaptation for Diffusion Models via Maximum Mean Discrepancy Guidance

arXiv:2601.08379v2 Announce Type: replace-cross Abstract: Pre-trained diffusion models have emerged as powerful generative priors for both unconditional and conditional sample generation, yet their ou

MODIP: Efficient Model-Based Optimization for Diffusion Policies

arXiv:2606.10825v1 Announce Type: new Abstract: Diffusion policies (DPs) have emerged as expressive policy representations for robot learning, often used with imitation learning methods such as behavi

Multi-Faceted Interactivity Alignment in Full-Duplex Speech Models

arXiv:2606.11167v1 Announce Type: new Abstract: Full-duplex spoken dialogue models can listen and speak simultaneously, making them a promising architecture for natural conversation. However, current

Multilingual Word-Level Forced Alignment with Self-Supervised Representations and Learned Dynamic Programming

arXiv:2606.10675v1 Announce Type: new Abstract: We present a method for accurate multilingual word-level forced alignment, consisting of an alignment encoder and a learned alignment decoder. The encod

New series from The Observer with @Yoshua_Bengio, me, and others on potential catastrophic risks in AI. (link at bottom of thread)

New series from The Observer with @Yoshua_Bengio, me, and others on potential catastrophic risks in AI. (link at bottom of thread) For the last few months I’ve been investigating how concerned we shou

Offline Reinforcement Learning for Rotation Profile Control in Tokamaks

arXiv:2605.05857v2 Announce Type: replace Abstract: Tokamaks remain leading candidates for achieving practical fusion energy, yet many important control problems inside these devices are still difficu

On-sky demonstration of reinforcement learning for adaptive optics control

arXiv:2606.10771v1 Announce Type: cross Abstract: Reinforcement learning (RL)-based algorithms have recently emerged as a promising approach for adaptive optics (AO) control. In simulations and labora

One paper from three years ago has been influencing policymakers around the world about labour decisions regarding AI. What happens when the…

One paper from three years ago has been influencing policymakers around the world about labour decisions regarding AI. What happens when the gap between the evidence and new policy widens? Our report

safetycohere--x

$ORCL hit hard after hours on plan for raising even more money for AI fairytales. And this is despite the fact that [see CNBC headline on ri…

Gary Marcus critiques Oracle's after-hours stock decline following announcements of increased capital allocation toward AI initiatives, suggesting skepticism about the company's AI strategy and spendi

PADD: Path-Aligned Decompression Distillation for Non-Router Teacher to Guide MoE Student Learning

arXiv:2606.10369v1 Announce Type: new Abstract: As large language models (LLMs) continue to scale, it becomes increasingly challenging to grow model capacity under fixed computation budgets. We propos

Pareto-Guided Teacher Alignment for Fair Personalized Text Generation

arXiv:2606.10126v1 Announce Type: cross Abstract: Personalized persuasive text generation can improve relevance and engagement, but demographic conditioning may also introduce unequal framing across g

Model Releases

PreAct-Bench: Benchmarking Predictive Monitoring in LLMs

model-releasesarxiv-cs-ai

arXiv:2606.09890v1 Announce Type: cross Abstract: Large language models (LLMs) are increasingly deployed as autonomous agents capable of executing multi-step action trajectories toward a given objecti

Prediction (which I have made before and which increasingly seems likely) OpenAI is gonna fall, and take a lot down with it.

Prediction (which I have made before and which increasingly seems likely) OpenAI is gonna fall, and take a lot down with it. BREAKING: SoftBank shares fall 9% after its attempt secure a $6 billion mar

Question: tech market is down about 10% over the last week. Will that affect SpaceX’s IPO? Or will everything proceed as planned, since it i…

A question posed on X regarding whether a recent 10% decline in the tech market will impact SpaceX's planned IPO timing and execution. The post, from AI researcher and entrepreneur Gary Marcus, raises

really good point. who else?

really good point. who else? BREAKING; Bill Gates just told Congress that Jeffrey Epstein had sought to use Gates’ affairs in an effort to blackmail him. So the question now is, if he did this to Bill

Reasoning or Memorization? Direction-Aware Diversity Exploration in LLM Reinforcement Learning

arXiv:2606.10346v1 Announce Type: new Abstract: Reinforcement learning has become a key paradigm for eliciting reasoning abilities in large language models, where exploration is crucial for discoverin

Reasoning over Semantic IDs Enhances Generative Recommendation

arXiv:2603.23183v2 Announce Type: replace-cross Abstract: Recent advances in generative recommendation have leveraged pretrained LLMs by formulating sequential recommendation as autoregressive generat

Recovering the Zipfian Distribution in Unsupervised Term Discovery

arXiv:2606.10781v1 Announce Type: cross Abstract: Unsupervised term discovery involves segmenting unlabelled speech into word- or syllable-like units and clustering these into a lexicon of candidate t