safety

Beyond Attack Success Rate: Temporal Logit Observability for LLM Safety Failures

arXiv:2605.29629v1 Announce Type: new Abstract: Attack Success Rate (ASR) evaluates each jailbreak with a single yes/no label at the end of generation, telling us whether a failure happened but not ho

Human

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Safety

Beyond Bilingual Transfer: Multilingual Code-Switching in Instruction Tuning

DGX agent

arXiv:2605.29414v1 Announce Type: cross Abstract: Recent studies have shown that code-switching data (CSD), in which multiple languages are mixed within the same context, can improve cross-lingual tra

safetyarxiv-cs-ai

29 May 2026

Beyond Trajectory Rewards: Step-level Credit Assignment for Agentic Search via Graph Modeling

arXiv:2605.29697v1 Announce Type: new Abstract: In Agentic Search, trajectory-level outcome rewards fail to quantify the behavioral contributions of individual steps, while existing step-level reward

bold set of counterpredictions, from @scaling01:

bold set of counterpredictions, from @scaling01: Cold take on what comes next: - OpenAI will flourish - Anthropic will continue to be profitable - Google will not catch up to Anthropic or OpenAI - no

BORA: Bridging Offline Reinforcement Learning and Online Residual Adaptation for Real-World Dexterous VLA Models

arXiv:2605.30226v1 Announce Type: cross Abstract: Vision-Language-Action (VLA) models have emerged as a promising paradigm for grounding visual-language understanding into real-world robotic manipulat

Bridging the Sim-to-Real Gap in Reinforcement Learning-Based Industrial Dispatching through Execution Semantics

arXiv:2605.29078v1 Announce Type: new Abstract: Event-driven scheduling policies are increasingly deployed in industrial environments, where decisions are made under asynchronous and partially observe

Calibration Is Not Enough: Evaluating Confidence Estimation Under Language Variations

arXiv:2601.08064v2 Announce Type: replace Abstract: Confidence estimation (CE) indicates how reliable the answers of large language models are and impacts user trust and decision-making. Existing eval

Causal Interventions on Continuous Variables: A Case Study on Verb Bias in Steering Vectors for In-Context Learning

arXiv:2605.29971v1 Announce Type: new Abstract: Causal interventions in language model representations have largely targeted discrete features, like grammatical number. However, language models must a

Causal-JEPA: Learning World Models through Object-Level Latent Masking

arXiv:2602.11389v2 Announce Type: replace Abstract: World models require robust relational understanding to support prediction, reasoning, and control. While object-centric representations provide a u

CB-SLICE: Concept-Based Interpretable Error Slice Discovery

arXiv:2605.29836v1 Announce Type: cross Abstract: Despite strong average-case performance, deep learning models often exhibit systematic errors on specific population groups, known as error slices. Id

Certified Policy Optimisation for Nested Causal Bandits via PAC-Bayes Risk

arXiv:2605.29788v1 Announce Type: new Abstract: Critical sequential decisions are rarely single-timescale: a strategic decision causally shapes the context in which every subsequent tactical choice is

Colored Noise Diffusion Sampling

arXiv:2605.30332v1 Announce Type: new Abstract: Diffusion models achieve state-of-the-art image synthesis, with their generative trajectories fundamentally exhibiting a spectral bias, resolving low-fr

Comparative evaluation of photogrammetric reconstruction methods and 3D Gaussian Splatting for road surface roughness analysis

arXiv:2605.29452v1 Announce Type: new Abstract: Image-based 3D reconstruction offers a low-cost alternative to traditional sensor-based techniques for road surface assessment. This study compares four

Crafting Desirable Climate Trajectories with RL Explored Socio-Environmental Simulations

arXiv:2410.07287v2 Announce Type: replace-cross Abstract: Climate change poses an existential threat, necessitating effective climate policies to enact impactful change. Decisions in this domain are i

CRITIC-R1: Learning Structured Critics for Retrieval-Augmented Generation

arXiv:2605.29886v1 Announce Type: cross Abstract: Retrieval-augmented generation (RAG) improves knowledge-intensive question answering by incorporating external evidence. However, existing RAG methods

Cycle Consistency in Video Object-Centric Learning

arXiv:2605.30211v1 Announce Type: new Abstract: Self-supervised video Object-Centric Learning (OCL) aims to discover distinct objects and associate them across time, whereas self-supervised Multi-Obje

DAMEL: Dual-Axis Multi-Expert Learning for Class-Imbalanced Learning

arXiv:2605.30135v1 Announce Type: cross Abstract: Various algorithms have been proposed to address the challenges posed by class-imbalanced learning from real-world data with long-tailed distributions

DeepSurvey: Enhancing Analytical Depth and Citation Reliability in Automated Survey Generation

arXiv:2605.29522v1 Announce Type: new Abstract: As scientific literature grows rapidly, automated survey generation has become a key capability for AI scientists and human researchers. However, existi

DefSynUS: Real-time Patient-specific Intrahepatic Vessel Identification via Deformation-Aware CT-US Domain Adaptation

arXiv:2605.29570v1 Announce Type: new Abstract: Purpose: Laparoscopic ultrasound (LUS) enhances the safety of liver surgery by visualizing intrahepatic vessels in real-time. Still, vessel identificati

Deja View: Looping Transformers for Multi-View 3D Reconstruction

arXiv:2605.30215v1 Announce Type: new Abstract: Recent feed-forward 3D reconstruction transformers have scaled to over a billion parameters, following the broader trend of increasing model capacity in

Discovering Cooperative Pipelines: Autoresearch for Sequential Social Dilemmas

arXiv:2605.30003v1 Announce Type: cross Abstract: We study two-level autoresearch for cooperation: an outer-loop AI agent autonomously redesigns the inner-loop pipeline of an LLM policy-synthesis syst

DLM-SWAI: Steering Diffusion Language Models Before They Unmask

arXiv:2605.29626v1 Announce Type: cross Abstract: Steering language model generation toward desired textual properties is essential for practical deployment, and inference-time methods are particularl

Do Deep Networks Forget Initialization? A Forgetting-Time View of Practical Inductive Bias

arXiv:2605.29152v1 Announce Type: new Abstract: Randomly initialized neural networks induce a prior over functions, but the predictor used in practice is produced only after training. We ask how much

Draft-OPD: On-Policy Distillation for Speculative Draft Models

arXiv:2605.29343v1 Announce Type: new Abstract: Speculative decoding accelerates large language model inference by pairing a target model with a lightweight draft model whose proposed tokens are verif

Dual-Stream Diffusion for World-Model Augmented Vision-Language-Action Model

arXiv:2510.27607v3 Announce Type: replace Abstract: Augmenting vision-language-action models (VLAs) with world models is promising for robotic policy learning but faces challenges in jointly predictin

DynaFLIP: Rethinking Robotics Perception via Tri-Modal-Dynamics Guided Representation

arXiv:2605.30350v1 Announce Type: cross Abstract: Robot manipulation critically depends on perception that preserves the action-relevant aspects of a scene. Yet most robot learning pipelines are built

Dynamics Within Latent Chain-of-Thought: An Empirical Study of Causal Structure

arXiv:2602.08783v3 Announce Type: replace Abstract: Latent or continuous chain-of-thought methods replace explicit textual rationales with a number of internal latent steps, but these intermediate com

EAPO: Enhancing Policy Optimization with On-Demand Expert Assistance

arXiv:2509.23730v2 Announce Type: replace Abstract: Large language models (LLMs) have recently advanced in reasoning when optimized with reinforcement learning (RL) under verifiable rewards. Existing

Emergent Semantic Representations in World Models through Physical Interaction without Linguistic Supervision

arXiv:2605.28865v1 Announce Type: cross Abstract: What does a world model learn from physical exploration, without any linguistic supervision? We argue the answer is organized by a single principle: t

Entropy-KL Divergence-based Token Masking: A Novel Approach for Selective Fine-tuning of Large Language Models

arXiv:2605.29303v1 Announce Type: new Abstract: Supervised fine-tuning (SFT) followed by reinforcement learning (RL) has become a standard post-training paradigm for large language models. This paradi

EPiC: Efficient Video Camera Control Learning with Precise Anchor-Video Guidance

arXiv:2505.21876v2 Announce Type: replace-cross Abstract: Recent approaches for video generation with camera control often create anchor videos (i.e., rendered videos that approximate desired camera m

even if @scaling01 turns out to be wrong about some of these, I respect the specificity.

even if @scaling01 turns out to be wrong about some of these, I respect the specificity. a bit more specific: - OpenAI will flourish -> meaning they will stay at the frontier and their market cap cont

Evolutionary Refinement of Generative Graph Topologies: A Hybrid WGAN-GA Approach

arXiv:2605.29161v1 Announce Type: cross Abstract: Generating realistic graph-structured data is challenging due to discrete connectivity, varying graph sizes, and class-specific structural patterns. R

EvoMD-LLM: Learning the Language of Species Evolution in Reactive Molecular Dynamics

arXiv:2605.29394v1 Announce Type: new Abstract: While large language models (LLMs) excel at static scientific reasoning, they struggle to model the temporal structure of dynamic physical processes. We

EvoRubric: Self-Evolving Rubric-Driven RL for Open-Ended Generation

arXiv:2605.29847v1 Announce Type: new Abstract: Reinforcement Learning (RL) has significantly advanced Large Language Models (LLMs) in verifiable domains, but aligning models for open-ended generation

exactly this.

exactly this. @Michael14kBall @GaryMarcus @Vivek4real_ He's been publicly trashed by AI boosters this whole time, in dismissive terms. And the thing he's doing, with a number of others, is to try to c

FakeVLM-R1: Internalizing Physical Laws via CoT for Synthetic Image Detection

arXiv:2605.30062v1 Announce Type: new Abstract: The development of generative artificial intelligence technologies has propelled the visual realism of synthetic images to an unprecedented level. Altho

Fewer Steps, Better Performance: Efficient Cross-Modal Clip Trimming for Video Moment Retrieval Using Language

arXiv:2605.29793v1 Announce Type: new Abstract: Given an untrimmed video and a sentence query, video moment retrieval using language (VMR) aims to locate a target query-relevant moment. Since the untr

Fisher-Preserving Guidance: Training-Free Manifold Constraints for Safe Diffusion Control

arXiv:2605.29937v1 Announce Type: cross Abstract: Diffusion models are effective for waypoint prediction in visual navigation, but standard sampling and test time guidance can produce unreliable or in

FlowSeg: Dynamic Semantic Guidance for LLM-Conditioned Segmentation

arXiv:2605.29461v1 Announce Type: new Abstract: LLM-conditioned segmentation has recently advanced rapidly by coupling large language models with iterative mask generation frameworks. However, we iden

Follow-the-Perturbed-Leader for Decoupled Bandits: Best-of-Both-Worlds and Practicality

arXiv:2510.12152v2 Announce Type: replace-cross Abstract: We study the decoupled multi-armed bandit problem, where the learner separately selects one arm for exploration and one, possibly different, a

Former Tesla data labelers say FSD relies on laborious mapping for hazards; crash data analysis shows Tesla exaggerates FSD's safety via flawed methodology (Reuters)

Reuters: Former Tesla data labelers say FSD relies on laborious mapping for hazards; crash data analysis shows Tesla exaggerates FSD's safety via flawed methodology — Tesla says its Full Self-Driving

safetytechmeme

From Context Shift to Stylistic Collapse: Why Training Objectives Matter More Than Scale

arXiv:2605.28826v1 Announce Type: new Abstract: In modern LLMs, linguistic features function not as stylistic artifacts but as probes of probability mass, allocated under training alignment objectives

From General Vision to Reliable Traversability Estimation: Adapting Vision Foundation Models for Unstructured Outdoor Environments

arXiv:2605.29565v1 Announce Type: new Abstract: Vision-based approaches have become the dominant paradigm for traversability estimation in unstructured outdoor environments, typically adapting vision

fully agree!

fully agree! Artificial intelligences do not undergo experiences, do not possess a body, do not feel joy or pain, do not mature through relationships, and do not know from within what love, work, frie

Future Forcing: Future-aware Training-free KV Cache Policy for Autoregressive Video Generation

arXiv:2605.30083v1 Announce Type: new Abstract: Autoregressive (AR) video generation has emerged as a promising paradigm for long-horizon video synthesis, where each frame is generated conditioned on

GAP3D: Generative Alignment of VLM Latents to Patch-Level Embeddings for 3D Generation

arXiv:2605.28995v1 Announce Type: new Abstract: Recent approaches integrating vision-language models (VLMs) as prompt encoders for generative model conditioning typically rely on expensive end-to-end

GAPD: Gold-Action Policy Distillation for Agentic Reinforcement Learning in Knowledge Base Question Answering

arXiv:2605.29584v1 Announce Type: new Abstract: Reinforcement learning (RL) is a natural fit for agentic knowledge base question answering (KBQA), where a model must issue executable actions, observe