safety

Bringing Order to Asynchronous SGD: Towards Optimality under Data-Dependent Delays with Momentum

arXiv:2605.02043v1 Announce Type: new Abstract: Asynchronous stochastic gradient descent (SGD) enables scalable distributed training but suffers from gradient staleness. Existing mitigation strategies

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Safety

Brockman confirms that Sam was fired for not being consistently candid. Crazy that @karaswisher blocked me for saying that the board fired S…

DGX agent

Brockman confirms that Sam was fired for not being consistently candid. Crazy that @karaswisher blocked me for saying that the board fired Sam for not being consistently candid, when that is in fact w

safetygary-marcus--x

5 May 2026

Brockman’s counsel is doing a good job of laying out the timeline — but done little so far to refute yesterday’s dissection of her client’s …

Brockman’s counsel is doing a good job of laying out the timeline — but done little so far to refute yesterday’s dissection of her client’s self-dealing and dodgy behavior regarding his fiduciary resp

safetygary-marcus--x

Causal Foundations of Collective Agency

arXiv:2605.00248v1 Announce Type: new Abstract: A key challenge for the safety of advanced AI systems is the possibility that multiple simpler agents might inadvertently form a collective agent with c

safetyarxiv-cs-ai

Colinearity Decay: Training Quantization-Friendly ViTs with Outlier Decay

arXiv:2605.01330v1 Announce Type: new Abstract: Low-bit quantization is a practical route for efficiently deploying vision Transformers, yet activation outliers complicate fully quantized deployment.

Combining Facial Videos and Biosignals for Stress Estimation During Driving

arXiv:2601.04376v3 Announce Type: replace Abstract: Reliable stress recognition is critical in applications such as medical monitoring and safety-critical systems, including real-world driving. While

Combining Trained Models in Reinforcement Learning

arXiv:2605.02159v1 Announce Type: new Abstract: Deep reinforcement learning (DRL) has delivered strong results in domains such as Atari and Go, but it still suffers from high sample cost and weak tran

Compared to What? Baselines and Metrics for Counterfactual Prompting

arXiv:2605.01048v1 Announce Type: new Abstract: Counterfactual prompting (i.e., perturbing a single factor and measuring output change) is widely used to evaluate things like LLM bias and CoT faithful

Compliance-Aware Agentic Payments on Stablecoin Rails

arXiv:2605.00071v1 Announce Type: cross Abstract: Agentic payment systems extend delegated action to financial transfers, but scaling them on stablecoin rails in regulated settings requires safeguards

safetyarxiv-cs-ai

Confident, Calibrated, or Complicit: Safety Alignment and Ideological Bias in LLM Hate Speech Detection

arXiv:2509.00673v2 Announce Type: replace Abstract: We investigate the efficacy of Large Language Models (LLMs) in detecting implicit and explicit hate speech, examining how models with minimal safety

Contrastive Residual Energy Test-time Adaptation

arXiv:2505.19607v2 Announce Type: replace Abstract: Test-time adaptation (TTA) enhances model robustness by enabling adaptation to target distributions that differ from training distributions, improvi

CUE: Concept-Aware Multi-Label Expansion to Mitigate Concept Confusion in Long-Tailed Learning

arXiv:2605.01309v1 Announce Type: new Abstract: Long-tailed distributions are common in real-world recognition tasks, where a few head classes have many samples while most tail classes have very few.

Cut-In Gap Acceptance Toward Autonomous vs. Human-Driven Vehicles: Evidence from the Waymo Open Motion Dataset

arXiv:2605.01485v1 Announce Type: new Abstract: Autonomous vehicles (AVs) are widely known to follow conservative, rule-based motion policies that surrounding drivers can learn to anticipate. A direct

CycleRL: Sim-to-Real Deep Reinforcement Learning for Robust Autonomous Bicycle Control

arXiv:2603.15013v2 Announce Type: replace Abstract: Autonomous bicycles offer a promising agile solution for urban mobility and last-mile logistics. However, conventional control strategies often stru

Decision Boundary-aware Generation for Long-tailed Learning

arXiv:2605.01468v1 Announce Type: new Abstract: Long-tailed data bias decision boundaries toward head classes and degrade tail class accuracy. Diffusion-based generative augmentation address this prob

DeepStage: Learning Autonomous Defense Policies Against Multi-Stage APT Campaigns

arXiv:2603.16969v2 Announce Type: replace-cross Abstract: This paper presents DeepStage, a deep reinforcement learning (DRL) framework for adaptive and stage-aware defense against Advanced Persistent

Delayed homomorphic reinforcement learning for environments with delayed feedback

arXiv:2604.03641v2 Announce Type: replace Abstract: Reinforcement learning in real-world systems often involves delayed feedback, which breaks the Markov assumption and impedes both learning and contr

Differential Parity: Relative Fairness Between Two Sets of Decisions

arXiv:2112.11279v4 Announce Type: replace Abstract: With AI systems widely applied to assist humans in decision-making processes such as talent hiring, school admission, and loan approval; there is an

Disciplined Diffusion: Text-to-Image Diffusion Model against NSFW Generation

arXiv:2605.01113v1 Announce Type: new Abstract: Text-to-image (T2I) diffusion models have the ability to build high-quality pictures from text prompts, but they pose safety concerns because they can g

Divide and Conquer: Decoupled Representation Alignment for Multimodal World Models

arXiv:2605.01896v1 Announce Type: new Abstract: Emerging multi-modal world models attempt to jointly generate videos across diverse modalities (e.g., RGB, depth, and mask), yet they fail to fully expl

Do Large Language Models Plan Answer Positions? Position Bias in Multiple-Choice Question Generation

arXiv:2605.01846v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly used to generate multiple-choice questions (MCQs), where correct answers should ideally be uniformly distr

DR-SNE: Density-Regularized Stochastic Neighbor Embedding

arXiv:2605.02060v1 Announce Type: new Abstract: Dimensionality reduction methods such as t-SNE are designed to preserve local neighborhood structure but do not explicitly account for how probability m

Dynamics Aware Quadrupedal Locomotion via Intrinsic Dynamics Head

arXiv:2605.01227v1 Announce Type: new Abstract: Quadrupedal locomotion plays a critical role in enabling agile, versatile movement across complex terrains. Understanding and estimating the underlying

Dynamics Distillation for Efficient and Transferable Control Learning

arXiv:2605.01516v1 Announce Type: new Abstract: Robust control policy learning for autonomous driving requires training environments to be both physically realistic and computationally scalable, prope

Edge Case Detection in Automated Driving: Methods, Challenges and Future Directions

arXiv:2410.08491v2 Announce Type: replace Abstract: Automated vehicles promise to enhance transportation safety and efficiency. However, ensuring their reliability in real-world conditions remains cha

Elon: Let’s settle. Greg: Nope. Elon: Ok, let’s talk about your diaries, then.

Elon: Let’s settle. Greg: Nope. Elon: Ok, let’s talk about your diaries, then. On the eve of trial, Elon Musk reached out to Greg Brockman about a potential settlement, according to court documents fi

safetygary-marcus--x

Evidence-Based Landing Site Selection and Vison-Based Landing for UAVs in Unstructured Environments

arXiv:2605.01432v1 Announce Type: new Abstract: Autonomous landing in cluttered or unstructured environments remains a safety-critical challenge for unmanned aerial vehicles (UAVs), particularly under

Experience Constrained Hierarchical Federated Reinforcement Learning for Large-scale UAV Teams in Hazardous Environments

arXiv:2605.02165v1 Announce Type: new Abstract: Conventional federated learning assumes that greater learner participation improves training performance, by leveraging abundant, independently generate

Exploring Data-Free LoRA Transferability for Video Diffusion Models

arXiv:2605.01929v1 Announce Type: new Abstract: Video diffusion models leveraging step distillation or causal distillation have achieved remarkable performance. However, adapting existing LoRAs to the

Exploring Entropy-based Active Learning for Fair Brain Segmentation

arXiv:2605.01706v1 Announce Type: new Abstract: Active learning (AL) has emerged as a crucial strategy for reducing the prohibitive costs associated with medical image segmentation. However, standard

Exploring Prompt Alignment with Clinical Factors in Zero-Shot Segmentation VLMs for NSCLC Tumor Segmentation

arXiv:2605.01266v1 Announce Type: new Abstract: Zero-shot vision-language models (VLMs) offer a promptable alternative to task-specific training for gross tumor volume (GTV) delineation in non-small-c

ExpoCM: Exposure-Aware One-Step Generative Single-Image HDR Reconstruction

arXiv:2605.02464v1 Announce Type: new Abstract: Single-image HDR reconstruction aims to recover high dynamic range radiance from a single low dynamic range (LDR) input, but remains highly ill-posed du

FastDSAC: Unlocking the Potential of Maximum Entropy RL in High-Dimensional Humanoid Control

arXiv:2603.12612v2 Announce Type: replace Abstract: Scaling Maximum Entropy Reinforcement Learning (RL) to high-dimensional humanoid control remains a fundamental challenge, as the ''curse of dimensio

Fine-Grained Class-Conditional Distribution Balancing for Debiased Learning

arXiv:2505.06831v2 Announce Type: replace Abstract: Achieving group-robust generalization in the presence of spurious correlations remains a significant challenge, particularly when bias annotations a

FLoRA: Fusion-Latent for Optical Reconstruction and Flood Area Segmentation via Cross-Modal Multi-Task Distillation Network

arXiv:2605.02137v1 Announce Type: new Abstract: Accurate flood water mapping is critical for disaster management, yet current methods struggle to fully exploit the potential of spaceborne imagery. Opt

Folks, AFAIK this is literally not a possible outcome of the trial, as Elon has waived the right to any cash payment to him, instead assig…

Folks, AFAIK this is *literally not a possible outcome of the trial*, as Elon has waived the right to any cash payment to him, instead assigning any monetary damages to OpenAI’s nonprofit. @PursueOpti

safetygary-marcus--x

Foundation Models to Unlock Real-World Evidence from Nationwide Medical Claims

arXiv:2605.02740v1 Announce Type: cross Abstract: Evidence derived from large-scale real-world data (RWD) is increasingly informing regulatory evaluation and healthcare decision-making. Administrative

From Concept to Capability: Evaluating 3D Gaussian Splatting for Synthetic Scene Editing in Autonomous Driving

arXiv:2605.01995v1 Announce Type: new Abstract: The perception of an Autonomous Driving System (ADS) critically depends on relevant, comprehensive, and diverse datasets to ensure its safety while oper

From Static Constraints to Dynamic Adaptation: Sample-Level Constraint Release for Offline-to-Online Reinforcement Learning

arXiv:2511.03828v2 Announce Type: replace Abstract: Offline-to-online reinforcement learning (O2O RL) faces a central challenge between retaining offline conservatism and adapting to online feedback u

General Frameworks for Conditional Two-Sample Testing

arXiv:2410.16636v2 Announce Type: replace-cross Abstract: We study the problem of conditional two-sample testing, which aims to determine whether two populations have the same distribution after accou

Generalized Distributional Alignment Games for Unbiased Answer-Level Fine-Tuning

arXiv:2605.02435v1 Announce Type: new Abstract: The Distributional Alignment Game framework provides a powerful variational perspective on Answer-Level Fine-Tuning (ALFT). However, standard algorithms

Geometric and Spectral Alignment for Deep Neural Network I

arXiv:2605.02108v1 Announce Type: new Abstract: Deep residual architectures are modeled as products of near-identity Jacobians. This paper proves deterministic quotient-geometric estimates for singula

Geometric and Spectral Alignment for Deep Neural Network II

arXiv:2605.02111v1 Announce Type: new Abstract: This paper develops the angular and static-channel component of Geometric and Spectral Alignment for residual Jacobian chains. Starting from Cartan-coor

GETA-3DGS: Automatic Joint Structured Pruning and Quantization for 3D Gaussian Splatting

arXiv:2605.02086v1 Announce Type: new Abstract: 3D Gaussian splatting (3DGS) is a state-of-the-art representation for real-time photorealistic novel-view synthesis, yet a single high-fidelity scene ty

Good in Bad (GiB): Sifting Through End-user Demonstrations for Learning a Better Policy

arXiv:2605.01529v1 Announce Type: new Abstract: Imitation learning offers a promising framework for enabling robots to acquire diverse skills from human users. However, most imitation learning algorit

Gradient-Gated DPO: Stabilizing Preference Optimization in Language Models

arXiv:2605.02626v1 Announce Type: new Abstract: Preference optimization has become a central paradigm for aligning large language models with human feedback. Direct Preference Optimization (DPO) simpl

Green Energy Management for Sustainable Data Centers Using Deep Reinforcement Learning

arXiv:2507.21153v2 Announce Type: replace Abstract: The exponential growth of digital services has positioned data centers among the most energy-intensive infrastructures in the modern economy, raisin

HandelBot: Real-World Piano Playing via Fast Adaptation of Dexterous Robot Policies

arXiv:2603.12243v3 Announce Type: replace Abstract: Mastering dexterous manipulation with multi-fingered hands has been a grand challenge in robotics for decades. Despite its potential, the difficulty