safety

Filter, Then Reweight: Rethinking Optimization Granularity in On-Policy Distillation

arXiv:2606.02684v1 Announce Type: cross Abstract: On-Policy distillation (OPD) in large language models is shifting from full-trace KL supervision toward more selective training paradigms. Recent OPD

Human

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Safety

First it was MIT and McKinsey. Now Bain finds that returns to corporate AI investments are disappointing.

DGX agent

A major consulting firm (Bain) has found that corporate returns on AI investments are underwhelming, following similar findings from MIT and McKinsey. This suggests that despite significant spending a

safetygary-marcus--x

3 Jun 2026

Flow Learners for PDEs: Toward a Physics-to-Physics Paradigm for Scientific Computing

arXiv:2604.07366v2 Announce Type: replace Abstract: Partial differential equations (PDEs) govern nearly every physical process in science and engineering, but solving them at scale remains prohibitive

Follow-Your-Preference++: Rethinking Preference Alignment for Image Inpainting

arXiv:2606.03216v1 Announce Type: new Abstract: We study preference alignment for image inpainting. Rather than proposing yet another method, we revisit the problem from first principles and reassess

Forgetting is Not Erasure: Recovering Latent Knowledge via Transport Keys

arXiv:2606.02860v1 Announce Type: cross Abstract: Catastrophic forgetting is often framed as a representational problem: after sequential training, a model appears to lose the features that supported

FreeStreamGS: Online Feed-forward 3D Gaussian Splatting from Unposed Streaming Inputs

arXiv:2606.03254v1 Announce Type: new Abstract: Feed-forward 3D Gaussian Splatting (3DGS) allows efficient and high-fidelity novel view synthesis (NVS) from an offline recorded image sequence. However

GeoAlign: Beyond Semantics with State-Guided Spatial Alignment in VLA Models

arXiv:2606.03240v1 Announce Type: new Abstract: Current Vision--Language--Action (VLA) models often optimize for semantic grounding, whereas executable manipulation requires geometry-aware spatial ali

safetyarxiv-cs-ro

GFFMERGE: Efficient Merging of Graph Neural Force Fields and Beyond

arXiv:2606.03232v1 Announce Type: cross Abstract: Graph Neural Networks (GNNs) have revolutionized Neural Force Fields for atomistic simulations, achieving near-quantum accuracy at reduced cost, yet a

Glass Box at Orbit: A Constitutional AI Verification Framework for Trustworthy Autonomous CubeSat Intelligence

arXiv:2606.02967v1 Announce Type: cross Abstract: The space industry is quietly building toward something nobody has fully reckoned with: orbital data centers running thousands of autonomous AI worklo

GLINT: Sparsely Gated Vision-Language Alignment for Fine-Grained Radiology Representations

arXiv:2606.03180v1 Announce Type: cross Abstract: Vision-language models (VLMs) for radiology have emerged as a scalable paradigm by leveraging image-report pairs naturally produced in clinical workfl

Grasp-Then-Plan with Failure Attribution: A Closed Two-Stage Framework for Precise and Generalizable Robotic Manipulation

arXiv:2606.03385v1 Announce Type: cross Abstract: In robotic manipulation, the tight coupling between grasping and motion planning often obscures the true source of failure, leading to inefficient tri

hard to make all these numbers square

hard to make all these numbers square 🦔IBM CEO Arvind Krishna says AI is 'not a bubble' then estimates the industry needs 6 to 8 trillion in total capex for data center and chip buildout. To recover t

safetygary-marcus--x

HARVE: Hacking-Aware Reward-Head Vector Editing for Robust Reward Models

arXiv:2606.03131v1 Announce Type: new Abstract: Reward models are central to large language model (LLM) alignment, but they remain vulnerable to reward hacking. To evaluate reward-model robustness, we

Hint-Guided Diversified Policy Optimization for LLM Reasoning

arXiv:2606.03021v1 Announce Type: new Abstract: Recent developments in Large Language Models (LLMs) have showcased impressive reasoning capabilities, with Reinforcement Learning with Verifiable Reward

How Kodiak trains the brain behind 28 driverless trucks

Twenty-eight trucks, and no humans in the cab. As of March 31, 2026, Kodiak's autonomous driving system, the Kodiak Driver, runs commercial freight on public roads across long-haul trucking, and indus

safetylambda-labs

How OpenAI, Anthropic, and AI startups are pursuing 'recursive self-improvement', in a bid to build AI that can improve itself with little to no human input (Financial Times)

Financial Times: How OpenAI, Anthropic, and AI startups are pursuing “recursive self-improvement”, in a bid to build AI that can improve itself with little to no human input — Industry chiefs say tech

safetytechmeme

Human-in-the-Loop Contextual Bandits for Short-Term Rental Dynamic Pricing: Structural Equivalence of Historical Warm-Up and Approval-Gated Live Learning

arXiv:2606.02595v1 Announce Type: new Abstract: Dynamic pricing in short-term rental (STR) markets presents a distinctive challenge for online learning algorithms: pricing decisions carry significant

I was there, sitting next to Sam Altman in the US Senate, when he said this wonderful thing about artists and creators. It was a beautiful s…

I was there, sitting next to Sam Altman in the US Senate, when he said this wonderful thing about artists and creators. It was a beautiful statement. In hindsight, though, I realized he was lying (und

safetygary-marcus--x

if you cheat on your taxes, you can get audited. but you are just a chump.

This post by AI researcher Gary Marcus likely critiques the disparity in tax enforcement, suggesting that individual taxpayers face audit risks for cheating while wealthy individuals or corporations e

safetygary-marcus--x

I’m thrilled to be joining the EU's AI Scientific Panel to advise on the implementation of the EU AI Act and help assess and address AI’s gr…

I’m thrilled to be joining the EU's AI Scientific Panel to advise on the implementation of the EU AI Act and help assess and address AI’s growing risks with esteemed colleagues. The AI Act, the EU's f

safetyyoshua-bengio--x

Impact of Graph Structure on Membership-Inference Risk for Graph Neural Networks

arXiv:2601.17130v2 Announce Type: replace Abstract: Graph neural networks (GNNs) are widely used for tasks such as node classification and link prediction, but their use in sensitive settings raises c

'Important You should give me full credits!': Exploring Prompt Injection Attacks on LLM-Based Automatic Grading Systems

arXiv:2606.03090v1 Announce Type: cross Abstract: The emergence of large language models (LLMs) has significantly accelerated recent research on LLM-based automatic grading (AG) systems. Benefiting fr

Inference Cost Attacks for Retrieval-Augmented Large Language Models

arXiv:2606.02643v1 Announce Type: cross Abstract: Retrieval-Augmented Generation (RAG)-enhanced LLM systems, while powerful, introduce substantial inference costs due to the inclusion of an extra mult

Inference-Time Scaling for Joint Audio-Video Generation

arXiv:2606.03183v1 Announce Type: cross Abstract: Joint audio-video generation aims to synthesize realistic audio-video pairs that are both semantically aligned with text prompts and precisely synchro

J'ai beaucoup apprécié mon récent passage à @Cdanslair pour discuter des risques de l'IA pour nos sociétés, nos économies, et nos démocratie…

J'ai beaucoup apprécié mon récent passage à @Cdanslair pour discuter des risques de l'IA pour nos sociétés, nos économies, et nos démocraties. Merci @Caroline_Roux pour l'invitation! https://www.youtu

safetyyoshua-bengio--x

Jailbreak Attack Initializations as Extractors of Compliance Directions

arXiv:2502.09755v4 Announce Type: replace-cross Abstract: Safety-aligned LLMs respond to prompts with either compliance or refusal, each corresponding to distinct directions in the model's activation

KBQA-R1: Reinforcing Large Language Models for Knowledge Base Question Answering

arXiv:2512.10999v3 Announce Type: replace Abstract: Knowledge Base Question Answering (KBQA) challenges models to bridge the gap between natural language and strict knowledge graph schemas by generati

KC-3DGS: Kurtosis-Constrained Gaussian Splatting for High-Fidelity View Synthesis

arXiv:2606.03120v1 Announce Type: new Abstract: 3D Gaussian Splatting (3DGS) enables real-time novel view synthesis by representing scenes as collections of anisotropic Gaussians optimized via differe

Language Models Need Sleep: Learning to Self-Modify and Consolidate Memories

arXiv:2606.03979v1 Announce Type: cross Abstract: The past few decades have witnessed significant advances in the design of machine learning algorithms, from early studies on task-specific shallow mod

LAP: An Agent-to-Instrument Protocol for Autonomous Science

arXiv:2606.03755v1 Announce Type: new Abstract: Autonomous science is moving from demonstration to infrastructure. Large language model agents now plan experiments, and self-driving laboratories execu

Large Language Models Are Overconfident in Their Own Responses

arXiv:2606.03437v1 Announce Type: new Abstract: Prior work has shown that instruction-tuned large language models (LLMs) are less well calibrated than their base pre-trained counterparts. However, lit

Latent Activation Editing: Inference-Time Refinement of Learned Policies for Safer Multirobot Navigation

arXiv:2509.20623v2 Announce Type: replace Abstract: Reinforcement learning has enabled significant progress in complex domains such as coordinating and navigating multiple quadrotors. However, even we

safetyarxiv-cs-ro

LC-SAC: Lyapunov-Constrained Soft Actor-Critic via Koopman Operator Theory for Trajectory Tracking and Stabilization

arXiv:2602.04132v4 Announce Type: replace-cross Abstract: Reinforcement Learning (RL) has achieved remarkable success in solving complex sequential decision-making problems. However, its application t

Learning Power Flow with Confidence: A Probabilistic Guarantee Framework for Voltage Risk

arXiv:2308.07867v4 Announce Type: replace-cross Abstract: The absence of formal performance guarantees in machine learning (ML) has limited its adoption for safety-critical power system applications,

Learning Self-Interpretation from Interpretability Artifacts: Training Lightweight Adapters on Vector-Label Pairs

arXiv:2602.10352v2 Announce Type: replace-cross Abstract: Self-interpretation methods prompt language models to describe their own internal states, but remain unreliable due to hyperparameter sensitiv

Learning to Bet for Horizon-Aware Anytime-Valid Testing

arXiv:2603.19551v2 Announce Type: replace-cross Abstract: We develop horizon-aware anytime-valid tests and confidence sequences for bounded means under a strict deadline N. Using the betting/e-process

Learning Unmasking Policies for Diffusion Language Models

arXiv:2512.09106v4 Announce Type: replace Abstract: Diffusion (Large) Language Models (dLLMs) now match the downstream performance of their autoregressive counterparts on many tasks, while holding the

Let There Be Light: Reflection, Refraction and Scattering for Neural Operators

arXiv:2606.03262v1 Announce Type: new Abstract: Neural operators learn mappings between infinite-dimensional function spaces and provide a data-driven surrogate modeling paradigm for parametric partia

Letting Tutor Personas Speak Up for LLMs: Learning Steering Vectors from Dialogue via Preference Optimization

arXiv:2602.07639v2 Announce Type: replace Abstract: With the emergence of large language models (LLMs) as a powerful class of generative artificial intelligence (AI), their use in tutoring has become

Leveraging BART to Assess CS1 C++ Programming Assignments using Rubric-based Criteria

arXiv:2606.03814v1 Announce Type: new Abstract: This paper investigates rubric-aware, multitask fine-tuning of transformer models for automated grading of introductory C++ programming assignments, wit

Libra: Efficient Resource Management for Agentic RL Post-Training

arXiv:2606.03077v1 Announce Type: cross Abstract: Reinforcement learning (RL) has become a standard post-training paradigm for large language models (LLMs), extending beyond preference alignment to co

Localized, High-resolution Geographic Representations with Slepian Functions

arXiv:2602.00392v2 Announce Type: replace Abstract: Geographic data is fundamentally local. Disease outbreaks cluster in population centers, ecological patterns emerge along coastlines, and economic a

LoCAtion: Long-time Collaborative Attention Framework for High Dynamic Range Video Reconstruction

arXiv:2603.14377v2 Announce Type: replace Abstract: Prevailing High Dynamic Range (HDR) video reconstruction methods are fundamentally trapped in a fragile alignment-and-fusion paradigm. While explici

Margin Play: A Multi-Agent System For Public Policy Analysis In The Brazilian Equatorial Margin

arXiv:2606.02614v1 Announce Type: cross Abstract: The Brazilian Equatorial Margin (BEM) is Brazil's next offshore oil frontier, with operations expected to begin in 2026 in the Foz do Amazonas basin.

Measuring Weak-to-Strong Legibility of Reasoning Models

arXiv:2603.20508v2 Announce Type: replace-cross Abstract: Reasoning language models (RLMs) and the intermediate chains of thought they emit play an increasingly central role in multi-agent setups such

MetaWorld: Scaling Multi-Agent Video World Model from Single-view Video Data

arXiv:2606.02753v1 Announce Type: cross Abstract: Video world models are a foundational generative technology for embodied AI and the Metaverse, yet existing approaches are inherently limited to a sin

MIND: Multi-rationale INtegrated Discriminative Reasoning Framework for Multi-modal Large Models

arXiv:2512.05530v2 Announce Type: replace Abstract: Recently, multimodal large language models (MLLMs) have been widely applied to reasoning tasks. However, they suffer from limited multi-rationale se

Mitigating False Credit Propagation: Probabilistic Graphical Reward Aggregation for Rubric-Based Reinforcement Learning

arXiv:2606.03361v1 Announce Type: new Abstract: Rubric-based rewards are increasingly used for open-ended language model post-training, but criterion-level scores are often aggregated as independent u