safety

Mesh Field Theory: Port-Hamiltonian Formulation of Mesh-Based Physics

arXiv:2605.00394v1 Announce Type: new Abstract: We present Mesh Field Theory (MeshFT) and its neural realization, MeshFT-Net: a structure-preserving framework for mesh-based continuum physics that cle

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Safety

Model-Based Reinforcement Learning with Double Oracle Efficiency in Policy Optimization and Offline Estimation

DGX agent

arXiv:2605.00393v1 Announce Type: new Abstract: Reinforcement learning (RL) in large environments often suffers from severe computational bottlenecks, as conventional regret minimization algorithms re

safetyarxiv-cs-lg

4 May 2026

New Mexico child safety trial: New Mexico asks a judge to declare Meta a public nuisance and to order it to pay $3.7B and overhaul its apps to protect children (Diana Novak Jones/Reuters)

Diana Novak Jones / Reuters: New Mexico child safety trial: New Mexico asks a judge to declare Meta a public nuisance and to order it to pay $3.7B and overhaul its apps to protect children — The U.S.

safetytechmeme

NEW paper from Sakana AI (ICLR 2026). A 7B Conductor model just hit SOTA on GPQA-Diamond and LiveCodeBench by orchestrating other LLMs inste…

NEW paper from Sakana AI (ICLR 2026). A 7B Conductor model just hit SOTA on GPQA-Diamond and LiveCodeBench by orchestrating other LLMs instead of solving problems itself. (great paper! bookmark it!) T

safetydair-ai--x

Online Self-Calibration Against Hallucination in Vision-Language Models

arXiv:2605.00323v1 Announce Type: new Abstract: Large Vision-Language Models (LVLMs) often suffer from hallucinations, generating descriptions that include visual details absent from the input image.

Optimal Spatio-Temporal Decoupling for Bayesian Conformal Prediction

arXiv:2605.00432v1 Announce Type: new Abstract: Online Conformal Prediction (CP) struggles to balance temporal adaptability and structural stability. Feedback-driven methods (e.g., Adaptive Conformal

Optimizing Resource-Constrained Non-Pharmaceutical Interventions for Multi-Cluster Outbreak Control Using Hierarchical Reinforcement Learning

arXiv:2603.19397v2 Announce Type: replace Abstract: Non-pharmaceutical interventions (NPIs), such as diagnostic testing and quarantine, are crucial for controlling infectious disease outbreaks but are

Persona-Grounded Safety Evaluation of AI Companions in Multi-Turn Conversations

arXiv:2605.00227v1 Announce Type: new Abstract: There are growing concerns about the risks posed by AI companion applications designed for emotional engagement. Existing safety evaluations often rely

PORTool: Importance-Aware Policy Optimization with Rewarded Tree for Multi-Tool-Integrated Reasoning

arXiv:2510.26020v2 Announce Type: replace Abstract: Multi-tool-integrated reasoning enables LLM-empowered tool-use agents to solve complex tasks by interleaving natural-language reasoning with calls t

Pose-Aware Diffusion for 3D Generation

arXiv:2605.00345v1 Announce Type: new Abstract: Generating pose-aligned 3D objects is challenging due to the spatial mismatches and transformation ambiguities inherent in decoupled canonical-then-rota

PPLLaVA: Varied Video Sequence Understanding With Prompt Guidance

arXiv:2411.02327v4 Announce Type: replace Abstract: In the past year, video-based large language models (Video LLMs) have achieved impressive progress, particularly in their ability to process long vi

PrefMoE: Robust Preference Modeling with Mixture-of-Experts Reward Learning

arXiv:2605.00384v1 Announce Type: new Abstract: Preference-based reinforcement learning offers a scalable alternative to manual reward engineering by learning reward structures from comparative feedba

safetyarxiv-cs-ro

Prompt-Induced Score Variance in Zero-Shot Binary Vision-Language Safety Classification

arXiv:2605.00326v1 Announce Type: new Abstract: Single-prompt first-token probabilities from zero-shot vision-language model (VLM) safety classifiers are treated as decision scores, but we show they a

Provable and scalable quantum Gaussian processes for quantum learning

arXiv:2605.00099v1 Announce Type: cross Abstract: Despite rapid recent advances in quantum machine learning, the field is in many ways stuck. Existing approaches can exhibit serious limitations, and w

Recovering Hidden Reward in Diffusion-Based Policies

arXiv:2605.00623v1 Announce Type: new Abstract: This paper introduces EnergyFlow, a framework that unifies generative action modeling with inverse reinforcement learning by parameterizing a scalar ene

safetyarxiv-cs-ro

Reinforcement Learning for LLM Post-Training: A Survey

arXiv:2407.16216v3 Announce Type: replace Abstract: Large language models (LLMs) trained via pretraining and supervised fine-tuning (SFT) can still produce harmful and misaligned outputs, or struggle

Reinforcement Learning with LLM-Guided Action Spaces for Synthesizable Lead Optimization

arXiv:2604.07669v2 Announce Type: replace Abstract: Lead optimization in drug discovery requires improving therapeutic properties while ensuring that molecular modifications correspond to feasible syn

Reinforcement Learning with Markov Risk Measures and Multipattern Risk Approximation

arXiv:2605.00654v1 Announce Type: new Abstract: For a risk-averse finite-horizon Markov Decision Problem, we introduce a special class of Markov coherent risk measures, called mini-batch measures. We

ReLay: Personalized LLM-Generated Plain-Language Summaries for Better Understanding, but at What Cost?

arXiv:2605.00468v1 Announce Type: new Abstract: Plain Language Summaries (PLS) aim to make research accessible to lay readers, but they are typically written in a one-size-fits-all style that ignores

ResRL: Boosting LLM Reasoning via Negative Sample Projection Residual Reinforcement Learning

arXiv:2605.00380v1 Announce Type: cross Abstract: Reinforcement Learning with Verifiable Rewards (RLVR) enhances reasoning of Large Language Models (LLMs) but usually exhibits limited generation diver

Resting Neurons, Active Insights: Robustify Activation Sparsity for Large Language Models

arXiv:2512.12744v3 Announce Type: replace Abstract: Activation sparsity offers a compelling route to accelerate large language model (LLM) inference by selectively suppressing hidden activations, yet

SAGA: Workflow-Atomic Scheduling for AI Agent Inference on GPU Clusters

arXiv:2605.00528v1 Announce Type: cross Abstract: AI agents execute tens to hundreds of chained LLM calls per task, yet GPU schedulers treat each call as independent, discarding gigabytes of intermedi

SAVGO: Learning State-Action Value Geometry with Cosine Similarity for Continuous Control

arXiv:2605.00787v1 Announce Type: new Abstract: While representation and similarity learning have improved the sample efficiency of Reinforcement Learning (RL), they are rarely used to shape policy up

Scaling Video Understanding via Compact Latent Multi-Agent Collaboration

arXiv:2605.00444v1 Announce Type: new Abstract: Multi-modal large language models (MLLMs) advance vision language understanding but face inherent limitations in long-video tasks due to bounded percept

SIMON: Saliency-aware Integrative Multi-view Object-centric Neural Decoding

arXiv:2605.00401v1 Announce Type: new Abstract: Recent EEG-to-image retrieval methods leverage pretrained vision encoders and foveation-inspired priors, but typically assume a fixed, center-focused vi

Soft Graph Diffusion Transformer for MIMO Detection

arXiv:2605.00449v1 Announce Type: cross Abstract: Learning-based MIMO detection has shown strong empirical performance, yet existing methods typically rely on fixed-depth architectures without explici

Soft-MSM: Differentiable Context-Aware Elastic Alignment for Time Series

arXiv:2605.00069v1 Announce Type: new Abstract: Elastic distances like dynamic time warping (DTW) are central to time series machine learning because they compare sequences under local temporal misali

Stable-GFlowNet: Toward Diverse and Robust LLM Red-Teaming via Contrastive Trajectory Balance

arXiv:2605.00553v1 Announce Type: new Abstract: Large Language Model (LLM) Red-Teaming, which proactively identifies vulnerabilities of LLMs, is an essential process for ensuring safety. Finding effec

Statistical Impossibility and Possibility of Aligning LLMs with Human Preferences: From Condorcet Paradox to Nash Equilibrium

arXiv:2503.10990v2 Announce Type: replace-cross Abstract: Aligning large language models (LLMs) with diverse human preferences is critical for ensuring fairness and informed outcomes when deploying th

The Determinism of Randomness: Latent Space Degeneracy in Diffusion Model

arXiv:2511.07756v4 Announce Type: replace Abstract: Diffusion models initialize generation from an isotropic Gaussian latent, yet changing only the random seed can substantially alter prompt faithfuln

Towards A Generative Protein Evolution Machine with DPLM-Evo

arXiv:2605.00182v1 Announce Type: new Abstract: Proteins are shaped by gradual evolution under biophysical and functional constraints. Protein language models learn rich evolutionary constraints from

Uniform-Correct Policy Optimization: Breaking RLVR's Indifference to Diversity

arXiv:2605.00365v1 Announce Type: cross Abstract: Reinforcement Learning with Verifiable Rewards (RLVR) has achieved substantial gains in single-attempt accuracy (Pass@1) on reasoning tasks, yet often

UniVidX: A Unified Multimodal Framework for Versatile Video Generation via Diffusion Priors

arXiv:2605.00658v1 Announce Type: new Abstract: Recent progress has shown that video diffusion models (VDMs) can be repurposed for diverse multimodal graphics tasks. However, existing methods often tr

Unlearning What Matters: Token-Level Attribution for Precise Language Model Unlearning

arXiv:2605.00364v1 Announce Type: new Abstract: Machine unlearning has emerged as a critical capability for addressing privacy, safety, and regulatory concerns in large language models (LLMs). Existin

Unlocking Zero-Shot Geospatial Reasoning via Indirect Rewards

arXiv:2510.00072v2 Announce Type: replace Abstract: Training robust reasoning vision-language models (VLMs) in rare domains (such as geospatial) is fundamentally constrained by supervision scarcity. W

Unpaired Image Deraining Using Reward-Guided Self-Reinforcement Strategy

arXiv:2605.00719v1 Announce Type: new Abstract: Unsupervised deraining has attracted attention for its ability to learn the real-world distribution of rain without paired supervision. However, the lac

VGR: Visual Grounded Reasoning

arXiv:2506.11991v3 Announce Type: replace-cross Abstract: In the field of multimodal chain-of-thought (CoT) reasoning, existing approaches predominantly rely on reasoning on pure language space, which

VLBiMan: Vision-Language Anchored One-Shot Demonstration Enables Generalizable Bimanual Robotic Manipulation

arXiv:2509.21723v4 Announce Type: replace Abstract: Achieving generalizable bimanual manipulation requires systems that can learn efficiently from minimal human input while adapting to real-world unce

safetyarxiv-cs-ro

VR founder Jaron Lanier says a future where people are paid for the data they give AI is better than one where everyone depends on billionai…

VR founder Jaron Lanier says a future where people are paid for the data they give AI is better than one where everyone depends on billionaires to survive 'artificial intelligence is a marketing term,

Wasserstein Distributionally Robust Regret Optimization for Reinforcement Learning from Human Feedback

arXiv:2605.00155v1 Announce Type: cross Abstract: Reinforcement learning from human feedback (RLHF) has become a core post-training step for aligning large language models, yet the reward signal used

Way stronger. Today was huge for Musk, bad for OpenAI.

Way stronger. Today was huge for Musk, bad for OpenAI. The case against OpenAI is getting markedly stronger now that Musk is off the stand. Why? Musk’s lawyer is interrogating OpenAI founder Greg Broc

Well-known, well-respected AI expert Stuart Russell is now testifying at the Musk-OpenAI trial – despite OpenAI’s BS attempts to exclude him…

Well-known, well-respected AI expert Stuart Russell is now testifying at the Musk-OpenAI trial – despite OpenAI’s BS attempts to exclude him. Read the backstory here: A new filing just dropped in the

What Physics do Data-Driven MoCap-to-Radar Models Learn?

arXiv:2605.00018v1 Announce Type: new Abstract: Data-driven MoCap-to-radar models generate plausible micro-Doppler spectrograms, but do they actually learn the underlying physics? We introduce a physi

Why do these influencers always say “just published” for papers published last year? This paper is great, and I have written extensively abo…

Why do these influencers always say “just published” for papers published last year? This paper is great, and I have written extensively about it in my newsletter. But c’mon. Apple didn’t “just publis

World Model for Robot Learning: A Comprehensive Survey

arXiv:2605.00080v1 Announce Type: cross Abstract: World models, which are predictive representations of how environments evolve under actions, have become a central component of robot learning. They s

Wow. Greg Brockman had a 10M side deal with Altman even in the early nonprofit days, which not disclosed to Elon or (I believe) in nonprofit…

Gary Marcus reported that Greg Brockman, OpenAI's President, had a $10 million side deal with Sam Altman during OpenAI's early nonprofit period that was not disclosed to Elon Musk or documented in non

A profile of BlackBerry's QNX division, whose operating system controls safety features in 275M cars and accounts for half of BlackBerry's revenue (Ben Cohen/Wall Street Journal)

Ben Cohen / Wall Street Journal: A profile of BlackBerry's QNX division, whose operating system controls safety features in 275M cars and accounts for half of BlackBerry's revenue — John Wall has spen

safetytechmeme

3 May 2026

agreed

agreed The core argument of my latest article 'There is no AI race' was that the 'AI race' between the US and China is a manufactured narrative concocted by US tech companies to serve their corporate