safety

OxEnsemble: Fair Ensembles for Low-Data Classification

arXiv:2512.09665v2 Announce Type: replace Abstract: We address the problem of fair classification in settings where data is scarce and unbalanced across demographic groups. Such low-data regimes are c

Part^{2}GS: Part-aware Modeling of Articulated Objects using 3D Gaussian Splatting

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

DGX agent

arXiv:2506.17212v2 Announce Type: replace Abstract: Articulated objects are common in the real world, yet modeling their structure and motion remains a challenging task for 3D reconstruction methods.

safetyarxiv-cs-cv

10 Apr 2026

Path Regularization: A Near-Complete and Optimal Nonasymptotic Generalization Theory for Multilayer Neural Networks and Double Descent Phenomenon

arXiv:2503.02129v2 Announce Type: replace-cross Abstract: Path regularization has shown to be a very effective regularization to train neural networks, leading to a better generalization property than

PEER: Unified Process-Outcome Reinforcement Learning for Structured Empathetic Reasoning

arXiv:2508.09521v2 Announce Type: replace Abstract: Emotional support conversations require more than fluent responses. Supporters need to understand the seeker's situation and emotions, adopt an appr

People in Washington get played, yet again We really should worry about cybersecurity - a lot – but Mythos is not the model these guys think…

People in Washington get played, yet again We really should worry about cybersecurity - a lot – but Mythos is not the model these guys think it is. (See my newsletter today for three reasons why it is

safetygary-marcus--x

Personalizing Text-to-Image Generation to Individual Taste

arXiv:2604.07427v1 Announce Type: new Abstract: Modern text-to-image (T2I) models generate high-fidelity visuals but remain indifferent to individual user preferences. While existing reward models opt

Plug-and-Play Logit Fusion for Heterogeneous Pathology Foundation Models

arXiv:2604.07779v1 Announce Type: new Abstract: Pathology foundation models (FMs) have become central to computational histopathology, offering strong transfer performance across a wide range of diagn

PolySLGen: Online Multimodal Speaking-Listening Reaction Generation in Polyadic Interaction

arXiv:2604.08125v1 Announce Type: new Abstract: Human-like multimodal reaction generation is essential for natural group interactions between humans and embodied AI. However, existing approaches are l

PriPG-RL: Privileged Planner-Guided Reinforcement Learning for Partially Observable Systems with Anytime-Feasible MPC

arXiv:2604.08036v1 Announce Type: cross Abstract: This paper addresses the problem of training a reinforcement learning (RL) policy under partial observability by exploiting a privileged, anytime-feas

Probabilistic Language Tries: A Unified Framework for Compression, Decision Policies, and Execution Reuse

arXiv:2604.06228v1 Announce Type: cross Abstract: We introduce probabilistic language tries (PLTs), a unified representation that makes explicit the prefix structure implicitly defined by any generati

Quality-preserving Model for Electronics Production Quality Tests Reduction

arXiv:2604.06451v1 Announce Type: new Abstract: Manufacturing test flows in high-volume electronics production are typically fixed during product development and executed unchanged on every unit, even

Qualixar OS: A Universal Operating System for AI Agent Orchestration

arXiv:2604.06392v1 Announce Type: new Abstract: We present Qualixar OS, the first application-layer operating system for universal AI agent orchestration. Unlike kernel-level approaches (AIOS) or sing

Reason in Chains, Learn in Trees: Self-Rectification and Grafting for Multi-turn Agent Policy Optimization

arXiv:2604.07165v1 Announce Type: new Abstract: Reinforcement learning for Large Language Model agents is often hindered by sparse rewards in multi-step reasoning tasks. Existing approaches like Group

Reason-SVG: Enhancing Structured Reasoning for Vector Graphics Generation with Reinforcement Learning

arXiv:2505.24499v2 Announce Type: replace Abstract: Generating high-quality Scalable Vector Graphics (SVGs) is challenging for Large Language Models (LLMs), as it requires advanced reasoning for struc

Reasoning Within the Mind: Dynamic Multimodal Interleaving in Latent Space

arXiv:2512.12623v3 Announce Type: replace-cross Abstract: Recent advancements in Multimodal Large Language Models (MLLMs) have significantly enhanced cross-modal understanding and reasoning by incorpo

Reflection-Based Task Adaptation for Self-Improving VLA

arXiv:2510.12710v3 Announce Type: replace Abstract: Pre-trained Vision-Language-Action (VLA) models represent a major leap towards general-purpose robots, yet efficiently adapting them to novel, speci

ReflectRM: Boosting Generative Reward Models via Self-Reflection within a Unified Judgment Framework

arXiv:2604.07506v1 Announce Type: cross Abstract: Reward Models (RMs) are critical components in the Reinforcement Learning from Human Feedback (RLHF) pipeline, directly determining the alignment qual

Region-R1: Reinforcing Query-Side Region Cropping for Multi-Modal Re-Ranking

arXiv:2604.05268v2 Announce Type: replace-cross Abstract: Multi-modal retrieval-augmented generation (MM-RAG) relies heavily on re-rankers to surface the most relevant evidence for image-question quer

Remember when ChatGPT helped someone plan a mass shooting, and a dozen OpenAI employees notified management, and they did nothing and 8 peop…

Remember when ChatGPT helped someone plan a mass shooting, and a dozen OpenAI employees notified management, and they did nothing and 8 people died? They don't want to be liable for their incompetence

safetygary-marcus--x

Reset-Free Reinforcement Learning for Real-World Agile Driving: An Empirical Study

arXiv:2604.07672v1 Announce Type: new Abstract: This paper presents an empirical study of reset-free reinforcement learning (RL) for real-world agile driving, in which a physical 1/10-scale vehicle le

Responsible and safe use of AI

The OpenAI Academy page on 'Responsible and Safe Use of AI' is a guidance resource focused on best practices for using ChatGPT responsibly in professional and personal settings. It emphasizes that ...

safetyopenai

Rethinking Generalization in Reasoning SFT: A Conditional Analysis on Optimization, Data, and Model Capability

arXiv:2604.06628v1 Announce Type: new Abstract: A prevailing narrative in LLM post-training holds that supervised finetuning (SFT) memorizes while reinforcement learning (RL) generalizes. We revisit t

Revisiting Fairness Impossibility with Endogenous Behavior

arXiv:2604.06378v1 Announce Type: cross Abstract: In many real-world settings, institutions can and do adjust the consequences attached to algorithmic classification decisions, such as the size of fin

RewardFlow: Generate Images by Optimizing What You Reward

arXiv:2604.08536v1 Announce Type: new Abstract: We introduce RewardFlow, an inversion-free framework that steers pretrained diffusion and flow-matching models at inference time through multi-reward La

RoboAgent: Chaining Basic Capabilities for Embodied Task Planning

arXiv:2604.07774v1 Announce Type: cross Abstract: This paper focuses on embodied task planning, where an agent acquires visual observations from the environment and executes atomic actions to accompli

RoSHI: A Versatile Robot-oriented Suit for Human Data In-the-Wild

arXiv:2604.07331v1 Announce Type: cross Abstract: Scaling up robot learning will likely require human data containing rich and long-horizon interactions in the wild. Existing approaches for collecting

Rotation Equivariant Convolutions in Deformable Registration of Brain MRI

arXiv:2604.08034v1 Announce Type: new Abstract: Image registration is a fundamental task that aligns anatomical structures between images. While CNNs perform well, they lack rotation equivariance - a

Safe Large-Scale Robust Nonlinear MPC in Milliseconds via Reachability-Constrained System Level Synthesis on the GPU

arXiv:2604.07644v1 Announce Type: new Abstract: We present GPU-SLS, a GPU-parallelized framework for safe, robust nonlinear model predictive control (MPC) that scales to high-dimensional uncertain rob

Seeing but Not Thinking: Routing Distraction in Multimodal Mixture-of-Experts

arXiv:2604.08541v1 Announce Type: cross Abstract: Multimodal Mixture-of-Experts (MoE) models have achieved remarkable performance on vision-language tasks. However, we identify a puzzling phenomenon t

Seeing Like an AI: How LLMs Apply (and Misapply) Wikipedia Neutrality Norms

arXiv:2407.04183v4 Announce Type: replace Abstract: Large language models (LLMs) are trained on broad corpora and then used in communities with specialized norms. Is providing LLMs with community rule

Self-Debias: Self-correcting for Debiasing Large Language Models

arXiv:2604.08243v1 Announce Type: new Abstract: Although Large Language Models (LLMs) demonstrate remarkable reasoning capabilities, inherent social biases often cascade throughout the Chain-of-Though

Self-Distilled RLVR

arXiv:2604.03128v2 Announce Type: replace Abstract: On-policy distillation (OPD) has become a popular training paradigm in the LLM community. This paradigm selects a larger model as the teacher to pro

Semantic-Aware UAV Command and Control for Efficient IoT Data Collection

arXiv:2604.08153v1 Announce Type: new Abstract: Unmanned Aerial Vehicles (UAVs) have emerged as a key enabler technology for data collection from Internet of Things (IoT) devices. However, effective d

Semantic Noise Reduction via Teacher-Guided Dual-Path Audio-Visual Representation Learning

arXiv:2604.08147v1 Announce Type: cross Abstract: Recent advances in audio-visual representation learning have shown the value of combining contrastive alignment with masked reconstruction. However, j

SeMoBridge: Semantic Modality Bridge for Efficient Few-Shot Adaptation of CLIP

arXiv:2509.26036v3 Announce Type: replace Abstract: While Contrastive Language-Image Pretraining (CLIP) excels at zero-shot tasks by aligning image and text embeddings, its performance in few-shot cla

Shortcut Learning in Glomerular AI: Adversarial Penalties Hurt, Entropy Helps

arXiv:2604.07936v1 Announce Type: new Abstract: Stain variability is a pervasive source of distribution shift and potential shortcut learning in renal pathology AI. We ask whether lupus nephritis glom

SIM1: Physics-Aligned Simulator as Zero-Shot Data Scaler in Deformable Worlds

arXiv:2604.08544v1 Announce Type: cross Abstract: Robotic manipulation with deformable objects represents a data-intensive regime in embodied learning, where shape, contact, and topology co-evolve in

SMPL-GPTexture: Dual-View 3D Human Texture Estimation using Text-to-Image Generation Models

arXiv:2504.13378v2 Announce Type: replace-cross Abstract: Generating high-quality, photorealistic textures for 3D human avatars remains a fundamental yet challenging task in computer vision and multim

Soft-Quantum Algorithms

arXiv:2604.06523v1 Announce Type: cross Abstract: Quantum operations on pure states can be fully represented by unitary matrices. Variational quantum circuits, also known as quantum neural networks, e

Spike-based alignment learning solves the weight transport problem

arXiv:2503.02642v3 Announce Type: replace-cross Abstract: In both machine learning and in computational neuroscience, plasticity in functional neural networks is frequently expressed as gradient desce

Sumo: Dynamic and Generalizable Whole-Body Loco-Manipulation

arXiv:2604.08508v1 Announce Type: new Abstract: This paper presents a sim-to-real approach that enables legged robots to dynamically manipulate large and heavy objects with whole-body dexterity. Our k

SymptomWise: A Deterministic Reasoning Layer for Reliable and Efficient AI Systems

arXiv:2604.06375v1 Announce Type: new Abstract: AI-driven symptom analysis systems face persistent challenges in reliability, interpretability, and hallucination. End-to-end generative approaches ofte

SYN-DIGITS: A Synthetic Control Framework for Calibrated Digital Twin Simulation

arXiv:2604.07513v1 Announce Type: cross Abstract: AI-based persona simulation -- often referred to as digital twin simulation -- is increasingly used for market research, recommender systems, and soci

Synthetic Data for any Differentiable Target

arXiv:2604.08423v1 Announce Type: new Abstract: What are the limits of controlling language models via synthetic training data? We develop a reinforcement learning (RL) primitive, the Dataset Policy G

Temporal Inversion for Learning Interval Change in Chest X-Rays

arXiv:2604.04563v2 Announce Type: replace-cross Abstract: Recent advances in vision--language pretraining have enabled strong medical foundation models, yet most analyze radiographs in isolation, over

The Art of (Mis)alignment: How Fine-Tuning Methods Effectively Misalign and Realign LLMs in Post-Training

arXiv:2604.07754v1 Announce Type: cross Abstract: The deployment of large language models (LLMs) raises significant ethical and safety concerns. While LLM alignment techniques are adopted to improve m

The Defense Trilemma: Why Prompt Injection Defense Wrappers Fail?

arXiv:2604.06436v2 Announce Type: cross Abstract: We prove that no continuous, utility-preserving wrapper defense-a function D: Xo X that preprocesses inputs before the model sees them-can make al

The End of the Foundation Model Era: Open-Weight Models, Sovereign AI, and Inference as Infrastructure

arXiv:2604.06217v1 Announce Type: cross Abstract: The foundation model era -- roughly 2020 to 2025 -- is over. The forces that defined it have inverted. Open source models have reached frontier perfor