Search: “safety”

Causal Direct Preference Optimization for Distributionally Robust Generative Recommendation

arXiv:2603.22335v2 Announce Type: replace-cross Abstract: Direct Preference Optimization (DPO) guides large language models (LLMs) to generate recommendations aligned with user historical behavior dis

Causal Machine Learning: A Survey and Open Problems

DGX agent

arXiv:2206.15475v3 Announce Type: replace Abstract: Causal Machine Learning (CausalML) is an umbrella term for machine learning methods that formalize the data-generation process as a structural causa

safetyarxiv-cs-lg

28 May 2026

CIRF: Tokenizing Chain-of-Thoughts into Reusable Functional Units for Efficient Latent Reasoning in Large Language Models

arXiv:2605.28292v1 Announce Type: new Abstract: Implicit Chain-of-Thought (CoT) reduces the inference cost of large language models by internalizing the explicit rationales. However, existing approach

CodeGENCAT: Generative Computerized Adaptive Testing for Open-ended Coding Problems

arXiv:2602.20020v2 Announce Type: replace Abstract: Existing Computerized Adaptive Testing (CAT) frameworks typically select questions based on the predicted likelihood that the student will answer co

Commit to the Bit: Reactive Reinforcement Learning Done Right

arXiv:2605.28276v1 Announce Type: new Abstract: Reinforcement learning algorithms are commonly analyzed (and designed) under the Markov assumption. This is unrealistic, as most environments encountere

Compositional Text-to-Image Generation Via Region-aware Bimodal Direct Preference Optimization

arXiv:2605.28615v1 Announce Type: new Abstract: Despite the rapid progress of text-to-image (T2I) models, generating images that accurately reflect complex compositional prompts (covering attribute bi

Con-DSO: Learning Short-Horizon Consistency Priors for RGB-D Direct Sparse Odometry

arXiv:2605.27952v1 Announce Type: new Abstract: Visual odometry (VO) is a fundamental component in robotics and augmented reality. RGB-D direct VO benefits from metric depth measurements, but it can d

Counterfactually Fair Regression via Optimal Transport

arXiv:2605.28251v1 Announce Type: cross Abstract: We consider the problem of learning a counterfactually fair regressor. We adopt a causal uncertainty view in which counterfactual fairness is defined

CPPO: Contrastive Perception Policy Optimization for VLM Agents

arXiv:2601.00501v2 Announce Type: replace Abstract: We introduce CPPO, a Contrastive Perception Policy Optimization method for finetuning vision--language models (VLMs). Reliable perception is a core

crazy that this was announced on the day tokenmaxxing died.

crazy that this was announced on the day tokenmaxxing died. Anthropic raised 65 billion in a funding round that valued the artificial intelligence company at 965 billion including the new investment,

Cross-Entropy Games and Frost Training

arXiv:2605.27701v1 Announce Type: new Abstract: We present Frost Training, a method for improving Monte Carlo-based policy optimization for a large family of LLM-as-a-judge tasks called Cross-Entropy

Cyberbullying Governance on Social Media: A Unified Framework from Content Identification to Intervention

arXiv:2605.27584v1 Announce Type: new Abstract: The proliferation of social media platforms and online communities has inadvertently catalyzed the spread of cyberbullying, hate speech, and other forms

DebFilter: Eradicating Biases Stashed in Value

arXiv:2605.28167v1 Announce Type: new Abstract: Text-to-image diffusion models, which are theoretically equivalent to score-based generative models, generate images through a multi-step denoising proc

Deconstructing Spatial Complexity: Hierarchical Decomposition for LLM Spatial Reasoning

arXiv:2605.28144v1 Announce Type: new Abstract: LLMs have shown remarkable proficiency in general language understanding and reasoning. However, they consistently underperform in spatial reasoning tha

DenoiseRL: Bootstrapping Reasoning Models to Recover from Noisy Prefixes

arXiv:2605.28421v1 Announce Type: new Abstract: Reinforcement learning has become a central paradigm for advancing reasoning in large language models, yet most existing methods still depend on stronge

Diagnosing Live Within-Policy Instruction Conflicts in LLM Agents with Witnessed Resolution Profiles

arXiv:2605.27784v1 Announce Type: new Abstract: LLM agents are governed by long-lived natural-language prompt policies, but individually reasonable standing rules can interact in uninspected ways. We

Did AI just solve math? Cal Newport podcast: https://youtu.be/fhZRWZ6J4k4?si=Gzut_5x856n5DW6o

Cal Newport discusses whether recent AI advances represent a breakthrough in mathematical problem-solving capabilities, examining the implications of AI systems' improved performance on complex mathem

Diffusion-Augmented Markov Decision Processes for Maximum Entropy Reinforcement Learning

arXiv:2512.02019v3 Announce Type: replace-cross Abstract: Diffusion models excel at sampling from complex, unnormalized distributions. In this work, we extend Maximum Entropy Reinforcement Learning (M

Diffusion Large Language Models for Visual Speech Recognition

arXiv:2605.28456v1 Announce Type: new Abstract: Existing Visual Speech Recognition (VSR) systems commonly rely on left-to-right autoregressive decoding, which can force premature decisions on visually

DiscoForcing: A Unified Framework for Real-Time Audio-Driven Character Control with Diffusion Forcing

arXiv:2605.28491v1 Announce Type: new Abstract: We study real-time audio-responsive character control as a deployment-faithful problem: strictly causal, bounded-latency streaming that must generate co

DREAM-R: Multimodal Speculative Reasoning with RL-Based Refined Drafting, Precise Verification, and Fully Parallel Execution

arXiv:2605.28678v1 Announce Type: new Abstract: Speculative reasoning has recently been proposed as a means to accelerate reasoning-intensive generation in large multimodal models, but its effectivene

EAPO: Entropy-Driven Adaptive Positive-Negative Sample Weighting for Policy Optimization in Open-Ended QA

arXiv:2605.27846v1 Announce Type: new Abstract: Large Reasoning Models are typically trained via reinforcement learning from verifiable rewards (RLVR). However, existing approaches adopt fixed weights

ECHO: Entropy-Confidence Hybrid Optimization for Test-Time Reinforcement Learning

arXiv:2602.02150v2 Announce Type: replace-cross Abstract: Test-time reinforcement learning generates multiple candidate answers via repeated rollouts and performs online updates using pseudo-labels co

Emerging Extrinsic Dexterity in Cluttered Scenes via Dynamics-aware Policy Learning

arXiv:2603.09882v2 Announce Type: replace-cross Abstract: Extrinsic dexterity leverages environmental contact to overcome the limitations of prehensile manipulation. However, achieving such dexterity

EntroAD: Structural Entropy-Guided Prompt Adaptation for Zero-Shot Anomaly Detection

arXiv:2605.28630v1 Announce Type: new Abstract: Zero-Shot Anomaly Detection (ZSAD) aims to detect anomalies in unseen domains without target-domain adaptation. Recent CLIP-based methods have shown pro

Escape the Language Prior: Mitigating Late-Stage Modality Collapse in Audio Reasoning via Modality-Aware Policy Optimization

arXiv:2605.27741v1 Announce Type: new Abstract: Audio and omni-modal large language models exhibit impressive cross-modal reasoning capabilities. However, applying standard reinforcement learning post

Evaluating the Realism of LLM-powered Social Agents: A Case Study of Reactions to Spanish Online News

arXiv:2605.28598v1 Announce Type: cross Abstract: LLM-powered social agents are increasingly used to simulate online social behavior, yet their realism remains difficult to validate. Existing work has

Examining Agents' Bias Amplification versus Suppression in Multi-Agent Systems

arXiv:2605.28098v1 Announce Type: new Abstract: Multi-agent systems are increasingly deployed to support various tasks where agents interact to achieve individual and collective objectives. Although t

F Scott Fitzgerald, writing about someone who may as well have been Peter Thiel: “They were careless people, Tom and Daisy- they smashed up …

F Scott Fitzgerald, writing about someone who may as well have been Peter Thiel: “They were careless people, Tom and Daisy- they smashed up things and creatures and then retreated back into their mone

FABSVer: Faster Training and Better Self-Verification for LLM Mathematical Reasoning

arXiv:2605.28389v1 Announce Type: new Abstract: While large language models have made significant progress in mathematical reasoning, they remain unreliable at judging the correctness of their own sol

FedEHR-Gen: Federated Synthetic Time-Series EHR Generation via Latent Space Alignment and Distribution-Aware Aggregation

arXiv:2605.27892v1 Announce Type: new Abstract: Synthetic Electronic Health Record (EHR) generation provides a promising avenue for data augmentation and cross-hospital modeling in privacy-constrained

From Affect to Complex Behavior: Advancing Multimodal Human-Centered AI at the 10th ABAW Workshop & Competition

arXiv:2605.27451v1 Announce Type: new Abstract: The 10th Affective & Behavior Analysis in-the-Wild (ABAW) Workshop and Competition, held at CVPR 2026, continues to advance research on modelling, analy

From AR to Diffusion: Efficiently Adapting Large Language Models with Strictly Causal and Elastic Horizons

arXiv:2605.27387v1 Announce Type: cross Abstract: Diffusion models promise efficient parallel text generation but rely on bidirectional attention, creating a structural mismatch with pre-trained Autor

From Learning Resources to Competencies: LLM-Based Tagging with Evidence and Graph Constraints

arXiv:2605.28483v1 Announce Type: new Abstract: Linking learning resources to a structured competency framework is key to enabling competency-based search and curriculum analytics in Learning Manageme

From Pixels to Words -- Towards Native One-Vision Models at Scale

arXiv:2605.28820v1 Announce Type: new Abstract: Current vision-language models (VLMs) typically stitch together separate image encoders and language decoders via multi-stage alignment, a modular frame

fun thread on consciousness with @Grimezsz:

Gary Marcus shares a discussion thread about consciousness with user @Grimezsz on X (formerly Twitter), likely exploring philosophical, scientific, or technical perspectives on consciousness and relat

GE-Sim 2.0: A Roadmap Towards Comprehensive Closed-loop Video World Simulators for Robotic Manipulation

arXiv:2605.27491v1 Announce Type: new Abstract: We introduce GE-Sim 2.0 (Genie Envisioner World Simulator 2.0), a closed-loop video world simulator for robotic manipulation. Building on the action-con

safetyarxiv-cs-ro

GeneralThinker: Domain-General Reasoning through Likelihood-Guided Answer-Conditioned Optimization

arXiv:2605.27934v1 Announce Type: new Abstract: Reinforcement learning with verifiable rewards improves language model reasoning, but its reliance on domain-specific verifiers, sparse outcome rewards,

Geometry of Human Perceptual Domains Emerges Transiently in LLM Representations

arXiv:2605.27970v1 Announce Type: new Abstract: While large language models (LLMs) are trained purely on textual data, prior work has shown that their internal representations can exhibit rich geometr

Geometry of Relaxed Fair Regression: A Unified Framework for Aware and Unaware Settings

arXiv:2605.28233v1 Announce Type: cross Abstract: Fairness-accuracy trade-offs are a central concern in the deployment of fairness-aware machine learning methods. When sensitive attributes are unavail

Grimlock: Guarding High-Agency Systems with eBPF and Attested Channels

arXiv:2605.27488v1 Announce Type: cross Abstract: Agentic systems increasingly run user-authored orchestration code that invokes tools, spawns subtasks, and delegates work across machines and clouds.

GS-FUSE: Granger-Supervised Gated Fusion and Multi-Granularity Alignment for Event-Driven Financial Forecasting

arXiv:2605.28520v1 Announce Type: new Abstract: Accurately forecasting the impact of salient financial events on markets is critical for investors and policymakers. However, existing multimodal time-s

Guaranteed Optimal Compositional Explanations for Neurons

arXiv:2511.20934v2 Announce Type: replace Abstract: Compositional explanations are a family of methods that aim to describe the spatial alignment between neurons' receptive field activations and conce

Heterogeneous Causal Discovery of Repeated Undesirable Health Outcomes

arXiv:2503.11477v2 Announce Type: replace Abstract: Understanding the factors that trigger or prevent undesirable health outcomes across patient subpopulations is essential for designing targeted inte

Hierarchical Synthetic Tabular Data Generation: A Hybrid Top-Down and Bottom-Up Framework

arXiv:2605.28198v1 Announce Type: new Abstract: Existing approaches for synthetic tabular data generation are based on either purely generative models or LLMs, both of which struggle with data heterog

HiRQA: Hierarchical Ranking and Quality Alignment for Opinion-Unaware Image Quality Assessment

arXiv:2508.15130v2 Announce Type: replace Abstract: Despite significant progress in no-reference image quality assessment (NR-IQA), dataset biases and reliance on subjective labels continue to hinder

Holy shit! They changed the rules for Elon again... They waved the profitability rule & are adding SpaceX to indices only 5 days after IPO..…

Holy shit! They changed the rules for Elon again... They waved the profitability rule & are adding SpaceX to indices only 5 days after IPO... normally it's 90 This forces 401k retirement & passive fun

How the Optimizer Shapes Learned Solutions in Equivariant Neural Networks

arXiv:2605.27662v1 Announce Type: cross Abstract: Equivariant neural networks encode geometric symmetries by construction, yet they are often difficult to optimize and can underperform less constraine