Search: “safety”

Inertia in Moral and Value Judgments of Large Language Models

arXiv:2408.09049v3 Announce Type: replace Abstract: Large Language Models (LLMs) behave non-deterministically, and prompting has become a common method for steering their outputs. A popular strategy i

Information Representation Fairness in Long-Document Embeddings: The Peculiar Interaction of Positional and Language Bias

DGX agent

arXiv:2601.16934v2 Announce Type: replace Abstract: To be discoverable in an embedding-based search process, each part of a document should be reflected in its embedding representation. To quantify an

safetyarxiv-cs-cl

21 Apr 2026

Instinct vs. Reflection: Unifying Token and Verbalized Confidence in Multimodal Large Models

arXiv:2604.17274v1 Announce Type: new Abstract: Multimodal Large Language Models (MLLMs) have demonstrated exceptional capabilities in various perception and reasoning tasks. Despite this success, ens

Inter-Agent Relative Representations for Multi-Agent Option Discovery

arXiv:2512.24827v3 Announce Type: replace Abstract: Temporally extended actions improve the ability to explore and plan in single-agent settings. In multi-agent settings, the exponential growth of the

IYKYK (But AI Doesn't): Automated Content Moderation Does Not Capture Communities' Heterogeneous Attitudes Towards Reclaimed Language

arXiv:2604.16654v1 Announce Type: new Abstract: Reclaimed slur usage is a common and meaningful practice online for many marginalized communities. It serves as a source of solidarity, identity, and sh

Last Wednesday, our founder and scientific advisor, @Yoshua_Bengio, was officially appointed an Officer of the Order of the British Empire (…

Last Wednesday, our founder and scientific advisor, @Yoshua_Bengio, was officially appointed an Officer of the Order of the British Empire (OBE). This prestigious distinction recognizes his contributi

safetyyoshua-bengio--x

LatentMimic: Terrain-Adaptive Locomotion via Latent Space Imitation

arXiv:2604.16440v1 Announce Type: new Abstract: Developing natural and diverse locomotion controllers for quadruped robots that can adapt to complex terrains while preserving motion style remains a si

safetyarxiv-cs-ro

Learning-Based Sparsification of Dynamic Graphs in Robotic Exploration Algorithms

arXiv:2604.16509v1 Announce Type: cross Abstract: Many robotic exploration algorithms rely on graph structures for frontier-based exploration and dynamic path planning. However, these graphs grow rapi

Less Noise, More Voice: Reinforcement Learning for Reasoning via Instruction Purification

arXiv:2601.21244v3 Announce Type: replace-cross Abstract: Reinforcement Learning with Verifiable Rewards (RLVR) has advanced LLM reasoning, but remains constrained by inefficient exploration under lim

LLM-Extracted Covariates for Clinical Causal Inference: Rethinking Integration Strategies

arXiv:2604.16763v1 Announce Type: new Abstract: Causal inference from electronic health records (EHR) is fundamentally limited by unmeasured confounding: critical clinical states such as frailty, goal

Mammo-FM: Breast-specific foundational model for Integrated Mammographic Diagnosis, Prognosis, and Reporting

arXiv:2512.00198v2 Announce Type: replace Abstract: Breast cancer is one of the leading causes of death among women worldwide. We introduce Mammo-FM, the first foundation model specifically for mammog

Mark Zuckerberg and Meta Platforms $META just sent a memo to employees saying Meta Platforms is installing a new tracking software on the co…

Mark Zuckerberg and Meta Platforms $META just sent a memo to employees saying Meta Platforms is installing a new tracking software on the computers of all employees in the United States 🇺🇸 so it can t

safetygary-marcus--x

MASPO: Unifying Gradient Utilization, Probability Mass, and Signal Reliability for Robust and Sample-Efficient LLM Reasoning

arXiv:2602.17550v3 Announce Type: replace Abstract: Existing Reinforcement Learning with Verifiable Rewards (RLVR) algorithms, such as GRPO, rely on rigid, uniform, and symmetric trust region mechanis

MASSIVE: 🇺🇸 The BBC just validated everything we've been saying. A clear pattern of trades right before major Trump announcements. Iran wa…

MASSIVE: 🇺🇸 The BBC just validated everything we've been saying. A clear pattern of trades right before major Trump announcements. Iran war. Tariff reversals. Policy shifts. We tracked a whale for wee

safetyyann-lecun--x

Mechanisms of Multimodal Synchronization: Insights from Decoder-Based Video-Text-to-Speech Synthesis

arXiv:2411.17690v3 Announce Type: replace-cross Abstract: Unified decoder-only transformers have shown promise for multimodal generation, yet the mechanisms by which they synchronize modalities with h

MESA: A Training-Free Multi-Exemplar Deep Framework for Restoring Ancient Inscription Textures

arXiv:2604.17390v1 Announce Type: new Abstract: Ancient inscriptions frequently suffer missing or corrupted regions from fragmentation, erosion, or other damage, hindering reading, and analysis. We re

Mix and Match: Context Pairing for Scalable Topic-Controlled Educational Summarisation

arXiv:2604.18087v1 Announce Type: new Abstract: Topic-controlled summarisation enables users to generate summaries focused on specific aspects of source documents. This paper investigates a data augme

Modeling User Exploration Saturation: When Recommender Systems Should Stop Pushing Novelty

arXiv:2604.16419v1 Announce Type: cross Abstract: Fairness-aware recommender systems often mitigate bias by increasing exposure to under-represented or long-tail content, commonly through mechanisms t

Motion-Guided Semantic Alignment with Negative Prompts for Zero-Shot Video Action Recognition

arXiv:2604.17062v1 Announce Type: new Abstract: Zero-shot action recognition is challenging due to the semantic gap between seen and unseen classes. We present a novel framework that enhances CLIP wit

Navigating Distribution Shifts in Medical Image Analysis: A Survey

arXiv:2411.05824v3 Announce Type: replace-cross Abstract: Medical Image Analysis (MedIA) has become indispensable in modern healthcare, enhancing clinical diagnostics and personalized treatment. Despi

Negative Advantage Is a Double-Edged Sword: Calibrating Advantage in GRPO for Deep Search

arXiv:2604.18235v1 Announce Type: new Abstract: Deep search agents can autonomously initiate multi-turn interactions with search engines, thereby exhibiting strong question-answering capabilities. Suc

OmniVLA-RL: A Vision-Language-Action Model with Spatial Understanding and Online RL

arXiv:2604.17706v1 Announce Type: new Abstract: Visual-Language-Action (VLA) models represent a paradigm shift in embodied AI, yet existing frameworks often struggle with imprecise spatial perception,

safetyarxiv-cs-ro

On the Convergence and Size Transferability of Continuous-depth Graph Neural Networks

arXiv:2510.03923v2 Announce Type: replace Abstract: Continuous-depth graph neural networks, also known as Graph Neural Differential Equations (GNDEs), combine the structural inductive bias of Graph Ne

On the Importance of Tactile Sensing for Imitation Learning: A Case Study on Robotic Match Lighting

arXiv:2504.13618v4 Announce Type: replace Abstract: The field of robotic manipulation has advanced significantly in recent years. At the sensing level, several novel tactile sensors have been develope

safetyarxiv-cs-ro

On the Shelf Life of Fine-Tuned LLM-Judges: Future-Proofing, Backward-Compatibility, and Question Generalization

arXiv:2509.23542v2 Announce Type: replace Abstract: The LLM-as-a-judge paradigm is widely used in both evaluating free-text model responses and reward modeling for model alignment and fine-tuning. Rec

Operationalizing Fairness in Text-to-Image Models: A Survey of Bias, Fairness Audits and Mitigation Strategies

arXiv:2604.16516v1 Announce Type: new Abstract: Text-to-Image (T2I) generation models have been widely adopted across various industries, yet are criticized for frequently exhibiting societal stereoty

OPSDL: On-Policy Self-Distillation for Long-Context Language Models

arXiv:2604.17535v1 Announce Type: new Abstract: Extending the effective context length of large language models (LLMs) remains a central challenge for real-world applications. While recent post-traini

OVOD-Agent: A Markov-Bandit Framework for Proactive Visual Reasoning and Self-Evolving Detection

arXiv:2511.21064v2 Announce Type: replace-cross Abstract: Open-Vocabulary Object Detection (OVOD) aims to enable detectors to generalize across categories by leveraging semantic information. Although

PaTaRM: Bridging Pairwise and Pointwise Signals via Preference-Aware Task-Adaptive Reward Modeling

arXiv:2510.24235v3 Announce Type: replace Abstract: Reward models (RMs) are central to reinforcement learning from human feedback (RLHF), providing the critical supervision signals that align large la

Peerispect: Claim Verification in Scientific Peer Reviews

arXiv:2604.17667v1 Announce Type: new Abstract: Peer review is central to scientific publishing, yet reviewers frequently include claims that are subjective, rhetorical, or misaligned with the submitt

PEPR: Privileged Event-based Predictive Regularization for Domain Generalization

arXiv:2602.04583v2 Announce Type: replace Abstract: Deep neural networks for visual perception are highly susceptible to domain shift, which poses a critical challenge for real-world deployment under

Plasticity Loss in Deep Reinforcement Learning: A Survey

arXiv:2411.04832v3 Announce Type: replace-cross Abstract: Plasticity refers to a network's ability to adapt to changing data distributions, which is crucial for the successful training of deep reinfor

Poetry overheard on http://B.sky:

Gary Marcus shared an observation or commentary about poetry encountered on a platform or service referenced as 'B.sky' (likely Bluesky, the decentralized social network). The post appears to document

safetygary-marcus--x

Policy Testing in Markov Decision Processes

arXiv:2505.15342v2 Announce Type: replace-cross Abstract: We study the policy testing problem in discounted Markov decision processes (MDPs) in the fixed-confidence setting under a generative model wi

PrinciplismQA: A Philosophy-Grounded Approach to Assessing LLM-Human Clinical Medical Ethics Alignment

arXiv:2508.05132v2 Announce Type: replace Abstract: As medical LLMs transition to clinical deployment, assessing their ethical reasoning capability becomes critical. While achieving high accuracy on k

ProtoCLIP: Prototype-Aligned Latent Refinement for Robust Zero-Shot Chest X-Ray Classification

arXiv:2604.18444v1 Announce Type: cross Abstract: Zero-shot vision-language models (VLMs) have shown promise for chest radiograph classification, but their performance is often limited by confounding

Q-SINDy: Quantum-Kernel Sparse Identification of Nonlinear Dynamics with Provable Coefficient Debiasing

arXiv:2604.16779v1 Announce Type: cross Abstract: Quantum feature maps offer expressive embeddings for classical learning tasks, and augmenting sparse identification of nonlinear dynamics (SINDy) with

RAYEN: Imposition of Hard Convex Constraints on Neural Networks

arXiv:2307.08336v2 Announce Type: replace Abstract: Despite the numerous applications of convex constraints in Robotics, enforcing them within learning-based frameworks remains an open challenge. Exis

Reasoning on the Manifold: Bidirectional Consistency for Self-Verification in Diffusion Language Models

arXiv:2604.16565v1 Announce Type: new Abstract: While Diffusion Large Language Models (dLLMs) offer structural advantages for global planning, efficiently verifying that they arrive at correct answers

RemoteShield: Enable Robust Multimodal Large Language Models for Earth Observation

arXiv:2604.17243v1 Announce Type: new Abstract: A robust Multimodal Large Language Model (MLLM) for Earth Observation should maintain consistent interpretation and reasoning under realistic input vari

Render-of-Thought: Rendering Textual Chain-of-Thought as Images for Visual Latent Reasoning

arXiv:2601.14750v3 Announce Type: replace Abstract: Chain-of-Thought (CoT) prompting has achieved remarkable success in unlocking the reasoning capabilities of Large Language Models (LLMs). Although C

Rethinking the Comparison Unit in Sequence-Level Reinforcement Learning: An Equal-Length Paired Training Framework from Loss Correction to Sample Construction

arXiv:2604.17328v1 Announce Type: new Abstract: This paper investigates the length problem in sequence-level relative reinforcement learning. We observe that, although existing methods partially allev

Retrieval-Augmented Multimodal Model for Fake News Detection

arXiv:2604.18112v1 Announce Type: new Abstract: In recent years, multimodal multidomain fake news detection has garnered increasing attention. Nevertheless, this direction presents two significant cha

Reward Score Matching: Unifying Reward-based Fine-tuning for Flow and Diffusion Models

arXiv:2604.17415v1 Announce Type: cross Abstract: Reward-based fine-tuning aims to steer a pretrained diffusion or flow-based generative model toward higher-reward samples while remaining close to the

REZE: Representation Regularization for Domain-adaptive Text Embedding Pre-finetuning

arXiv:2604.17257v1 Announce Type: new Abstract: Recent text embedding models are often adapted to specialized domains via contrastive pre-finetuning (PFT) on a naive collection of scattered, heterogen

Robust Tool Use via Fission-GRPO: Learning to Recover from Execution Errors

arXiv:2601.15625v2 Announce Type: replace Abstract: Large language models (LLMs) can call tools effectively, yet they remain brittle in multi-turn execution: after a tool-call error, smaller models of

S-GRPO: Unified Post-Training for Large Vision-Language Models

arXiv:2604.16557v1 Announce Type: cross Abstract: Current post-training methodologies for adapting Large Vision-Language Models (LVLMs) generally fall into two paradigms: Supervised Fine-Tuning (SFT)

Scalable Neighborhood-Based Multi-Agent Actor-Critic

arXiv:2604.18190v1 Announce Type: new Abstract: We propose MADDPG-K, a scalable extension to Multi-Agent Deep Deterministic Policy Gradient (MADDPG) that addresses the computational limitations of cen