safety

Pope Francis: Died within hours after meeting JD Vance Viktor Orban: Lost within a few days after meeting JD Vance Middle east peace negotia…

Pope Francis: Died within hours after meeting JD Vance Viktor Orban: Lost within a few days after meeting JD Vance Middle east peace negotiations: fell apart within 21 hours of Vance’s arrival Welcome

safetygary-marcus--x

Safety

Post-Hoc Guidance for Consistency Models by Joint Flow Distribution Learning

DGX agent

arXiv:2604.08828v1 Announce Type: cross Abstract: Classifier-free Guidance (CFG) lets practitioners trade-off fidelity against diversity in Diffusion Models (DMs). The practicality of CFG is however h

safetyarxiv-cs-cv

13 Apr 2026

Post-Selection Distributional Model Evaluation

arXiv:2603.23055v2 Announce Type: replace-cross Abstract: Formal model evaluation methods typically certify that a model satisfies a prescribed target key performance indicator (KPI) level. However, i

Predicting Metabolic Dysfunction-Associated Steatotic Liver Disease using Machine Learning Methods: A Retrospective Cohort Study

arXiv:2510.22293v4 Announce Type: replace Abstract: Background: Metabolic dysfunction-associated steatotic liver disease (MASLD) affects 30-40% of US adults and is the most common chronic liver diseas

RAMP: Hybrid DRL for Online Learning of Numeric Action Models

arXiv:2604.08685v1 Announce Type: new Abstract: Automated planning algorithms require an action model specifying the preconditions and effects of each action, but obtaining such a model is often hard.

Re-Mask and Redirect: Exploiting Denoising Irreversibility in Diffusion Language Models

arXiv:2604.08557v1 Announce Type: cross Abstract: Diffusion-based language models (dLLMs) generate text by iteratively denoising masked token sequences. We show that their safety alignment rests on a

Reducing Class Bias In Data-Balanced Datasets Through Hardness-Based Resampling

arXiv:2504.07031v2 Announce Type: replace Abstract: Class-bias, that is class-wise performance disparities, is typically attributed to data imbalance and addressed through frequency-based resampling.

Region-Constrained Group Relative Policy Optimization for Flow-Based Image Editing

arXiv:2604.09386v1 Announce Type: new Abstract: Instruction-guided image editing requires balancing target modification with non-target preservation. Recently, flow-based models have emerged as a stro

Reinforcement-aware Knowledge Distillation for LLM Reasoning

arXiv:2602.22495v2 Announce Type: replace-cross Abstract: Reinforcement learning (RL) post-training has recently driven major gains in long chain-of-thought reasoning large language models (LLMs), but

Rethinking Prospect Theory for LLMs: Revealing the Instability of Decision-Making under Epistemic Uncertainty

arXiv:2508.08992v3 Announce Type: replace Abstract: Prospect Theory (PT) models human decision-making behaviour under uncertainty, among which linguistic uncertainty is commonly adopted in real-world

SafeMind: A Risk-Aware Differentiable Control Framework for Adaptive and Safe Quadruped Locomotion

arXiv:2604.09474v1 Announce Type: cross Abstract: Learning-based quadruped controllers achieve impressive agility but typically lack formal safety guarantees under model uncertainty, perception noise,

Scene-Agnostic Object-Centric Representation Learning for 3D Gaussian Splatting

arXiv:2604.09045v1 Announce Type: new Abstract: Recent works on 3D scene understanding leverage 2D masks from visual foundation models (VFMs) to supervise radiance fields, enabling instance-level 3D s

Scheming in the wild: detecting real-world AI scheming incidents with open-source intelligence

arXiv:2604.09104v1 Announce Type: cross Abstract: Scheming, the covert pursuit of misaligned goals by AI systems, represents a potentially catastrophic risk, yet scheming research suffers from signifi

SCoRe: Clean Image Generation from Diffusion Models Trained on Noisy Images

arXiv:2604.09436v1 Announce Type: new Abstract: Diffusion models trained on noisy datasets often reproduce high-frequency training artifacts, significantly degrading generation quality. To address thi

Score-Driven Rating System for Sports

arXiv:2604.09143v1 Announce Type: new Abstract: This paper introduces a score-driven rating system, a generalization of the classical Elo rating system that employs the score, i.e. the gradient of the

Semantic Intent Fragmentation: A Single-Shot Compositional Attack on Multi-Agent AI Pipelines

arXiv:2604.08608v1 Announce Type: cross Abstract: We introduce Semantic Intent Fragmentation (SIF), an attack class against LLM orchestration systems where a single, legitimately phrased request cause

SHIFT: Steering Hidden Intermediates in Flow Transformers

arXiv:2604.09213v1 Announce Type: new Abstract: Diffusion models have become leading approaches for high-fidelity image generation. Recent DiT-based diffusion models, in particular, achieve strong pro

Sim-to-Real Transfer for Muscle-Actuated Robots via Generalized Actuator Networks

arXiv:2604.09487v1 Announce Type: cross Abstract: Tendon drives paired with soft muscle actuation enable faster and safer robots while potentially accelerating skill acquisition. Still, these systems

SPEAR: An Engineering Case Study of Multi-Agent Coordination for Smart Contract Auditing

arXiv:2602.04418v3 Announce Type: replace-cross Abstract: We present SPEAR, a multi-agent coordination framework for smart contract auditing that applies established MAS patterns in a realistic securi

SPPO: Sequence-Level PPO for Long-Horizon Reasoning Tasks

arXiv:2604.08865v1 Announce Type: new Abstract: Proximal Policy Optimization (PPO) is central to aligning Large Language Models (LLMs) in reasoning tasks with verifiable rewards. However, standard tok

SSPO: Subsentence-level Policy Optimization

arXiv:2511.04256v2 Announce Type: replace Abstract: As a key component of large language model (LLM) post-training, Reinforcement Learning from Verifiable Rewards (RLVR) has substantially improved rea

safetyarxiv-cs-cl

StaRPO: Stability-Augmented Reinforcement Policy Optimization

arXiv:2604.08905v1 Announce Type: new Abstract: Reinforcement learning (RL) is effective in enhancing the accuracy of large language models in complex reasoning tasks. Existing RL policy optimization

STCast: Adaptive Boundary Alignment for Global and Regional Weather Forecasting

arXiv:2509.25210v3 Announce Type: replace-cross Abstract: To gain finer regional forecasts, many works have explored the regional integration from the global atmosphere, e.g., by solving boundary equa

StructRL: Recovering Dynamic Programming Structure from Learning Dynamics in Distributional Reinforcement Learning

arXiv:2604.08620v1 Announce Type: cross Abstract: Reinforcement learning is typically treated as a uniform, data-driven optimization process, where updates are guided by rewards and temporal-differenc

SubQuad: Near-Quadratic-Free Structure Inference with Distribution-Balanced Objectives in Adaptive Receptor framework

arXiv:2602.17330v3 Announce Type: replace-cross Abstract: Comparative analysis of adaptive immune repertoires at population scale is hampered by two practical bottlenecks: the near-quadratic cost of p

Summary: AI Governance to Avoid Extinction

With AI capabilities rapidly increasing, humans appear close to developing AI systems that are better than human experts across all domains. This raises a series of questions about how the world will—

safetymiri

The causal relation between off-street parking and electric vehicle adoption in Scotland

arXiv:2604.09271v1 Announce Type: new Abstract: The transition to electric mobility hinges on maximising aggregate adoption while also facilitating equitable access. This study examines whether the 'c

The Hot Mess of AI: How Does Misalignment Scale With Model Intelligence and Task Complexity?

arXiv:2601.23045v2 Announce Type: replace Abstract: As AI becomes more capable, we entrust it with more general and consequential tasks. The risks from failure grow more severe with increasing task sc

The trend of treating all of AI as One Big Thing that always includes data centers & job changes & education changes & power & accelerating …

The trend of treating all of AI as One Big Thing that always includes data centers & job changes & education changes & power & accelerating science & misinformation & national security & corporate con

safetyethan-mollick--x

The Two-Stage Decision-Sampling Hypothesis: Understanding the Emergence of Self-Reflection in RL-Trained LLMs

arXiv:2601.01580v2 Announce Type: replace-cross Abstract: Self-reflection capabilities emerge in Large Language Models after RL post-training, with multi-turn RL achieving substantial gains over SFT c

'There is no advantage from being first. If you get first to a superintelligence you don't control, the superintelligence wins. Not the USA.…

'There is no advantage from being first. If you get first to a superintelligence you don't control, the superintelligence wins. Not the USA. Not China. Not the UK.' @andreamiotti of @ControlAI & TBC o

safetyconnor-leahy--x

Think Less, Know More: State-Aware Reasoning Compression with Knowledge Guidance for Efficient Reasoning

arXiv:2604.09150v1 Announce Type: new Abstract: Large Reasoning Models (LRMs) achieve strong performance on complex tasks by leveraging long Chain-of-Thought (CoT), but often suffer from overthinking,

safetyarxiv-cs-cl

Through Their Eyes: Fixation-aligned Tuning for Personalized User Emulation

arXiv:2604.09368v1 Announce Type: cross Abstract: Large language model (LLM) agents are increasingly deployed as scalable user simulators for recommender system evaluation. Yet existing simulators per

TME-PSR: Time-aware, Multi-interest, and Explanation Personalization for Sequential Recommendation

arXiv:2604.09439v1 Announce Type: cross Abstract: In this paper, we propose a sequential recommendation model that integrates Time-aware personalization, Multi-interest personalization, and Explanatio

Tora3: Trajectory-Guided Audio-Video Generation with Physical Coherence

arXiv:2604.09057v1 Announce Type: new Abstract: Audio-video (AV) generation has recently made strong progress in perceptual quality and multimodal coherence, yet generating content with plausible moti

Toward World Models for Epidemiology

arXiv:2604.09519v1 Announce Type: new Abstract: World models have emerged as a unifying paradigm for learning latent dynamics, simulating counterfactual futures, and supporting planning under uncertai

Towards Responsible Multimodal Medical Reasoning via Context-Aligned Vision-Language Models

arXiv:2604.08815v1 Announce Type: new Abstract: Medical vision-language models (VLMs) show strong performance on radiology tasks but often produce fluent yet weakly grounded conclusions due to over-re

Training event-based neural networks with exact gradients via Differentiable ODE Solving in JAX

arXiv:2603.08146v3 Announce Type: replace Abstract: Existing frameworks for gradient-based training of spiking neural networks face a trade-off: discrete-time methods using surrogate gradients support

Traj2Action: A Co-Denoising Framework for Trajectory-Guided Human-to-Robot Skill Transfer

arXiv:2510.00491v3 Announce Type: replace-cross Abstract: Learning diverse manipulation skills for real-world robots is severely bottlenecked by the reliance on costly and hard-to-scale teleoperated d

Truncated Rectified Flow Policy for Reinforcement Learning with One-Step Sampling

arXiv:2604.09159v1 Announce Type: new Abstract: Maximum entropy reinforcement learning (MaxEnt RL) has become a standard framework for sequential decision making, yet its standard Gaussian policy para

Unbiased Rectification for Sequential Recommender Systems Under Fake Orders

arXiv:2604.08550v1 Announce Type: cross Abstract: Fake orders pose increasing threats to sequential recommender systems by misleading recommendation results through artificially manipulated interactio

UniSemAlign: Text-Prototype Alignment with a Foundation Encoder for Semi-Supervised Histopathology Segmentation

arXiv:2604.09169v1 Announce Type: new Abstract: Semi-supervised semantic segmentation in computational pathology remains challenging due to scarce pixel-level annotations and unreliable pseudo-label s

VAG: Dual-Stream Video-Action Generation for Embodied Data Synthesis

arXiv:2604.09330v1 Announce Type: cross Abstract: Recent advances in robot foundation models trained on large-scale human teleoperation data have enabled robots to perform increasingly complex real-wo

Verbalizing LLMs' assumptions to explain and control sycophancy

arXiv:2604.03058v2 Announce Type: replace-cross Abstract: LLMs can be socially sycophantic, affirming users when they ask questions like 'am I in the wrong?' rather than providing genuine assessment.

Violence is not the answer. But maybe boycotts are?

Violence is not the answer. But maybe boycotts are? 🚨 NOW: The FBI is RAIDING the home of a 20-year-old man who threw a molotov cocktail at the home of OpenAI CEO Sam Altman Over a DOZEN federal agent

safetygary-marcus--x

VISOR: Agentic Visual Retrieval-Augmented Generation via Iterative Search and Over-horizon Reasoning

arXiv:2604.09508v1 Announce Type: cross Abstract: Visual Retrieval-Augmented Generation (VRAG) empowers Vision-Language Models to retrieve and reason over visually rich documents. To tackle complex qu

Visually-Guided Policy Optimization for Multimodal Reasoning

arXiv:2604.09349v1 Announce Type: cross Abstract: Reinforcement learning with verifiable rewards (RLVR) has significantly advanced the reasoning ability of vision-language models (VLMs). However, the

When & How to Write for Personalized Demand-aware Query Rewriting in Video Search

arXiv:2602.17667v2 Announce Type: replace-cross Abstract: In video search systems, user historical behaviors provide rich context for identifying search intent and resolving ambiguity. However, tradit