safety

NEW: Ahead of the SpaceX IPO, we tracked hundreds of promises that Musk has made over the years (FSD, Mars etc). His success rate is…not goo…

NEW: Ahead of the SpaceX IPO, we tracked hundreds of promises that Musk has made over the years (FSD, Mars etc). His success rate is…not good. And it’s getting worse. Some actual data in our months lo

safetygary-marcus--x

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Safety

Non-Uniform Noise-to-Signal Ratio in the REINFORCE Policy-Gradient Estimator

DGX agent

arXiv:2602.01460v3 Announce Type: replace-cross Abstract: Policy-gradient methods are widely used in reinforcement learning, yet training often becomes unstable or slows down as learning progresses. W

safetyarxiv-cs-lg

2 Jun 2026

NormEval: A Unified Multi-Metric Framework for Evaluating Semantic Fidelity in Text Normalization

arXiv:2511.20409v2 Announce Type: replace Abstract: Text normalization methods such as stemming and lemmatization are fundamental components of NLP pipelines. As new normalization tools are developed

safetyarxiv-cs-cl

Not convinced that this kind of nationalization by fiat is at all the right way to go (and for that matter don’t expect the current breed of…

Not convinced that this kind of nationalization by fiat is at all the right way to go (and for that matter don’t expect the current breed of technology to generate trillions), but I am glad that Sande

safetygary-marcus--x

ObjEmbed: Towards Universal Multimodal Object Embeddings

arXiv:2602.01753v3 Announce Type: replace Abstract: Aligning objects with corresponding textual descriptions is a fundamental challenge and a realistic requirement in vision-language understanding. Wh

Off-Policy Learning in Large Action Spaces: Optimization Matters More Than Estimation

arXiv:2509.03456v2 Announce Type: replace-cross Abstract: Off-policy evaluation (OPE) and off-policy learning (OPL) are foundational for decision-making in offline contextual bandits. Recent advances

On Effectiveness and Efficiency of Agentic Tool-calling and RL Training

arXiv:2606.00135v1 Announce Type: cross Abstract: Tool-calling is a central component of modern large language model (LLM) agents, equipping them with skills beyond their parametric knowledge. This pa

On the Limits of LLM Adaptability: Impact of Model-Internalized Priors on Annotation Task Performance

arXiv:2606.00467v1 Announce Type: cross Abstract: Large Language Models (LLMs) are increasingly used for zero-shot annotation and LLM-as-a-judge tasks, yet their reliability hinges on how model-intern

One Bias After Another: Mechanistic Reward Shaping and Persistent Biases in Language Reward Models

arXiv:2603.03291v2 Announce Type: replace-cross Abstract: Reward Models (RMs) are crucial for online alignment of language models (LMs) with human preferences. However, RM-based preference-tuning is v

OPD+: Rethinking the Advantage Design for On-Policy Distillation

arXiv:2606.01039v1 Announce Type: cross Abstract: On-policy distillation (OPD) is a widely used technique to transfer capabilities from capable teacher language models to the base student models, and

OpenAI says it has not donated to any super PACs and does not have an employee-funded PAC, and that Greg Brockman's support for Leading the Future is personal (OpenAI)

OpenAI: OpenAI says it has not donated to any super PACs and does not have an employee-funded PAC, and that Greg Brockman's support for Leading the Future is personal — AI is going to be one of the mo

safetytechmeme

Optimal Bayesian Stopping for Efficient Inference of Consistent LLM Answers

arXiv:2602.05395v2 Announce Type: replace-cross Abstract: A simple strategy for improving LLM accuracy, especially in math and reasoning problems, is to sample multiple responses and submit the answer

Optimizing Diversity and Quality through Base-Aligned Model Collaboration

arXiv:2511.05650v2 Announce Type: replace-cross Abstract: Alignment has greatly improved large language models (LLMs)' output quality at the cost of diversity, yielding highly similar outputs across g

OSCAR: Obstacle Survival Curves for Adaptive Robot Navigation

arXiv:2606.00990v1 Announce Type: new Abstract: A mobile robot following a graph of known routes can make costly navigation errors when a temporary obstacle blocks a critical edge: waiting too long be

PACE: Phase-Aware Chunk Execution for Robot Policies with Action Chunking

arXiv:2606.00537v1 Announce Type: new Abstract: Recent vision-language-action and diffusion-based robot policies often use action chunking, where each policy query predicts a sequence of future action

Paradoxical noise preference in RNNs

arXiv:2601.04539v2 Announce Type: replace-cross Abstract: In recurrent neural networks (RNNs) used to model biological neural networks, noise is typically introduced during training to emulate biologi

Partial Fairness Awareness: Belief-Guided Strategic Mechanism for Strategic Agents

arXiv:2606.00826v1 Announce Type: new Abstract: Strategic machine learning investigates scenarios where agents manipulate their features to receive favorable decisions from predictive models. To addre

Pave-GRPO: Beyond Instantaneous Guidance through Principled Average Velocity Decomposition

arXiv:2606.01636v1 Announce Type: new Abstract: Post-training via Group Relative Policy Optimization (GRPO) has emerged as a powerful paradigm for aligning flow-based generative models with human pref

Perspective on Bias in Biomedical AI: Preventing Downstream Healthcare Disparities

arXiv:2604.14514v2 Announce Type: replace Abstract: Healthcare disparities persist across socioeconomic boundaries, often attributed to unequal access to screening, diagnostics, and therapeutics. Howe

Perturbation Effects on Accuracy and Fairness among Similar Individuals

arXiv:2404.01356v3 Announce Type: replace-cross Abstract: Deep neural networks are vulnerable to adversarial perturbations that can simultaneously degrade prediction robustness and individual fairness

PHASOR: Phase-Anchored Universal Action Representations for Humanoid Embodiments

arXiv:2606.01851v1 Announce Type: new Abstract: Learning a good action embedding space is fundamental to scalable robot policy learning, yet existing methods treat action latents as task-specific inte

PhyScene3D: Physically Consistent Interactive 3D Tabletop Scene Generation

arXiv:2606.01649v1 Announce Type: new Abstract: Generating physically consistent 3D tabletop scenes is a fundamental yet underexplored problem for interactive and generalist robotic learning. The chal

Physically-Constrained Mamba-SDE for Remaining Useful Life Prediction under Irregular Observations

arXiv:2606.01894v1 Announce Type: new Abstract: Accurate Remaining Useful Life prediction is critical for industrial predictive maintenance. However, real-world deployment is challenging due to the ir

Policy and World Modeling Co-Training for Language Agents

arXiv:2606.02388v1 Announce Type: cross Abstract: Reinforcement learning (RL) improves large language model (LLM) agents by teaching them which actions lead to high rewards, but provides little superv

Policy-based Foveated Imaging and Perception

arXiv:2606.02565v1 Announce Type: new Abstract: Ultra-high-resolution image sensors offer the potential to capture fine spatial details critical for many visual perception tasks, but acquiring and pro

Pool-Select-Refine: Allocation-Aware Generative Dataset Distillation with Soft-Label-Guided Latent Refinement

arXiv:2606.01920v1 Announce Type: new Abstract: Diffusion-based dataset distillation has recently emerged as a promising paradigm for condensing large-scale datasets into compact synthetic sets. By le

Position: Beyond Sensitive Attributes, ML Fairness Should Quantify Structural Injustice via Social Determinants

arXiv:2508.08337v3 Announce Type: replace-cross Abstract: Algorithmic fairness research has largely framed unfairness as discrimination along sensitive attributes. However, this approach limits visibi

Position: Good Embodied Reward Models Need Bad Behavior Data

arXiv:2606.01036v1 Announce Type: new Abstract: This position paper argues that to obtain reliable embodied reward models, the community must invest in ``bad'' robot data: failed, suboptimal, error-pr

Position: Stop Preaching and Start Practising Data Frugality for Responsible Development of AI

arXiv:2602.19789v2 Announce Type: replace Abstract: This position paper argues that the machine learning community must move from preaching to practising data frugality for responsible artificial inte

Post-Deterministic Distributed Systems: A New Foundation for Trustworthy Autonomous Infrastructure

arXiv:2606.01722v1 Announce Type: cross Abstract: For decades, distributed systems have typically assumed that correct participants execute protocol-specified behavior with stable, externally defined,

PR2: Predictive Routing Replay for MoE-Based LLM Reinforcement Learning

arXiv:2606.00395v1 Announce Type: cross Abstract: Mixture of Experts (MoE) Large Language Models (LLMs) achieve strong performance at scale. However, reinforcement learning (RL) on MoE-based LLMs ofte

Principled Synthetic Data Enables the First Scaling Laws for LLMs in Recommendation

arXiv:2602.07298v3 Announce Type: replace-cross Abstract: Large Language Models (LLMs) represent a promising frontier for recommender systems, yet their development has been impeded by the absence of

Probabilistic Performance Guarantees for Multi-Task Reinforcement Learning

arXiv:2602.02098v2 Announce Type: replace-cross Abstract: Multi-task reinforcement learning trains generalist policies that can execute multiple tasks. While recent years have seen significant progres

Prospect-Theory Behavior from Bellman Optimality in MDPs with Catastrophic States

arXiv:2606.00970v1 Announce Type: new Abstract: We study risk-neutral control in Markov decision processes with an absorbing catastrophic state. Even though rewards are linear and the agent has no uti

Quantifying the Salience of Geo-Cultural Values for Pluralistic Safety Alignment

arXiv:2606.00369v1 Announce Type: cross Abstract: Safe global deployment of AI models requires alignment with human values that vary across cultures. Yet rater pools in safety evaluation datasets rema

Quantitative Movement Testing: Measuring Patient Movements from a Single Smartphone Video

arXiv:2606.02301v1 Announce Type: cross Abstract: Chronic pain diminishes quality of life by decreasing functional ability, yet objectively measuring this functional impact remains challenging in real

RADE: Random Add-Drop Edge as a Regularizer

arXiv:2606.00757v1 Announce Type: new Abstract: Graph Neural Networks (GNNs) suffer from overfitting and over-squashing of long-range information. Stochastic graph augmentations (e.g., edge deletion)

RAFT: Data Refinement and Adaptive Distillation for Domain Fine-Tuning with Alleviated Forgetting

arXiv:2606.00147v1 Announce Type: cross Abstract: Domain-specific supervised fine-tuning (SFT) often improves in-domain performance at the cost of degrading a model's general capabilities. We view thi

RAIGen: Rare Attribute Identification in Text-to-Image Generative Models

arXiv:2602.06806v2 Announce Type: replace Abstract: Text-to-image diffusion models achieve impressive generation quality but inherit and amplify training-data biases, skewing coverage of semantic attr

RankByGene: Gene-Guided Histopathology Representation Learning Through Cross-Modal Ranking Consistency

arXiv:2411.15076v3 Announce Type: replace-cross Abstract: Spatial transcriptomics (ST) provides essential spatial context by mapping gene expression within tissue, enabling detailed study of cellular

REAL: Resolving Knowledge Conflicts in Knowledge-Intensive Visual Question Answering via Reasoning-Pivot Alignment

arXiv:2602.14065v2 Announce Type: replace Abstract: Knowledge-intensive Visual Question Answering (KI-VQA) frequently suffers from severe knowledge conflicts caused by the inherent limitations of open

Real-Time Sensing of Inaccessible Physical Fields via an Edge-Deployable Hardware-Portable Graph Neural Operator

arXiv:2604.01802v2 Announce Type: replace Abstract: Real-time inference of inaccessible interior physical fields from sparse boundary observations is a fundamental but unresolved problem in scientific

REBot: From RAG to CatRAG with Semantic Enrichment and Graph Routing

arXiv:2510.01800v3 Announce Type: replace Abstract: Academic regulation advising is essential for helping students interpret and comply with institutional policies, yet building effective systems requ

Reconsidering Positional Supervision in Masked Diffusion Language Model Training

arXiv:2601.22947v2 Announce Type: replace Abstract: Masked diffusion language models (MDLMs) generate text by unmasking tokens in parallel and have recently emerged as alternatives to autoregressive l

safetyarxiv-cs-cl

RedDebate: Safer Responses Through Multi-Agent Red Teaming Debates

arXiv:2506.11083v3 Announce Type: replace Abstract: We introduce RedDebate, a novel multi-agent debate framework that provides the foundation for Large Language Models (LLMs) to identify and mitigate

safetyarxiv-cs-cl

ReFLEX: Length-Generalizable CSI Denoising for MIMO-OFDM via Relative-Frequency Bias

arXiv:2606.00263v1 Announce Type: cross Abstract: This letter studies CSI denoising for MIMO--OFDM with variable NR resource block (RB) allocations. ReFLEX is a length-generalizable Transformer whose

Regime-Adaptive Continual Learning for Portfolio Management

arXiv:2606.00143v1 Announce Type: cross Abstract: Financial markets are inherently non-stationary, exhibiting frequent regime shifts and structural changes that render traditional Portfolio Management

Regularized Offline Policy Optimization with Posterior Hybrid Bayesian Belief

arXiv:2606.00680v1 Announce Type: new Abstract: Offline reinforcement learning (RL) aims to optimize policies from pre-collected datasets. A bottleneck of this paradigm is managing epistemic uncertain