Search: “safety”

Verify-Gated Completion as Admission Control in a Governed Multi-Agent Runtime: A Bounded Architecture Case Study

arXiv:2605.17998v1 Announce Type: cross Abstract: As multi-agent systems move from short interactions to tool-using workflows with specialized roles and persistent state, completion becomes a runtime-

19 May 2026

Model Releases

Few-Shot Large Language Models for Actionable Triage Categorization of Online Patient Inquiries

DGX agent

arXiv:2605.15680v1 Announce Type: new Abstract: Online patient inquiries are often informal, incomplete, and written before professional assessment, yet they must still be routed to an appropriate lev

model-releasesarxiv-cs-cl

18 May 2026

MHGraphBench: Knowledge Graph-Grounded Benchmarking of Mental Health Knowledge in Large Language Models

arXiv:2605.15589v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly used in the mental health domain, yet it remains unclear how well they capture related biomedical knowledg

Optimizing LLM Inference: Fluid-Guided Online Scheduling with Memory Constraints

arXiv:2504.11320v3 Announce Type: replace-cross Abstract: Large language models now serve millions of users daily, with providers incurring costs exceeding $700,000 per day. Each request requires toke

Quantization Undoes Alignment: Bias Emergence in Compressed LLMs Across Models and Precision Levels

arXiv:2605.15208v1 Announce Type: cross Abstract: Large Language Models are routinely compressed via post-training quantization to reduce inference costs and memory footprint for cloud and edge deploy

These kids are serial criminals with a callous disregard for life. If they are ever released from jail they will surely harm again. Austin P…

model-releaseselon-musk--x

These kids are serial criminals with a callous disregard for life. If they are ever released from jail they will surely harm again. Austin PD, Travis Co. Sheriff Office & Manor PD did their job. Texas

CLOVER: Closed-Loop Value Estimation & Ranking for End-to-End Autonomous Driving Planning

arXiv:2605.15120v1 Announce Type: cross Abstract: End-to-end autonomous driving planners are commonly trained by imitating a single logged trajectory, yet evaluated by rule-based planning metrics that

Descriptor: Distance-Annotated Traffic Perception Question Answering (DTPQA)

arXiv:2511.13397v2 Announce Type: replace-cross Abstract: The remarkable progress of Vision-Language Models (VLMs) on a variety of tasks has raised interest in their application to automated driving.

Fusion-fission forecasts when AI will shift to undesirable behavior

arXiv:2605.14218v1 Announce Type: new Abstract: The key problem facing ChatGPT-like AI's use across society is that its behavior can shift, unnoticed, from desirable to undesirable -- encouraging self

Good to Go: The LOOP Skill Engine That Hits 99% Success and Slashes Token Usage by 99% via One-Shot Recording and Deterministic Replay

arXiv:2605.14237v1 Announce Type: new Abstract: Deploying AI agents for repetitive periodic tasks exposes a critical tension: Large Language Models (LLMs) offer unmatched flexibility in tool orchestra

NodeSynth: Socially Aligned Synthetic Data for AI Evaluation

arXiv:2605.14381v1 Announce Type: cross Abstract: Recent advancements in generative AI facilitate large-scale synthetic data generation for model evaluation. However, without targeted approaches, thes

Talk is (Not) Cheap: A Taxonomy and Benchmark Coverage Audit for LLM Attacks

arXiv:2605.15118v1 Announce Type: cross Abstract: We introduce a reusable framework for auditing whether LLM attack benchmarks collectively cover the threat surface: a 4imes6 Target imes Technique mat

The Moltbook Observatory Archive: an incremental dataset of agent-only social network activity

arXiv:2605.13860v1 Announce Type: cross Abstract: Moltbook is a social media platform in which posts and comments are authored exclusively by autonomous AI agents. We present the Moltbook Observatory

The most revealing thing about this AI leadership paper is that it reads less like a vision for innovation and more like a glossy whitepaper…

model-releasesyann-lecun--x

The most revealing thing about this AI leadership paper is that it reads less like a vision for innovation and more like a glossy whitepaper for a 21st century East India Company. Every generation of

Vision-Based Runtime Monitoring under Varying Specifications using Semantic Latent Representations

arXiv:2605.13923v1 Announce Type: cross Abstract: We study certified runtime monitoring of past-time signal temporal logic (ptSTL) from visual observations under partial observability. The monitor mus

Action Emergence from Streaming Intent

arXiv:2605.12622v1 Announce Type: cross Abstract: We formalize action emergence as a target capability for end-to-end autonomous driving: the ability to generate physically feasible, semantically appr

Connecting the Dots: A Machine Learning Ready Dataset for Ionospheric Forecasting Models

arXiv:2511.15743v2 Announce Type: replace Abstract: Operational forecasting of the ionosphere remains a critical space weather challenge due to sparse observations, complex coupling across geospatial

Negation Neglect: When models fail to learn negations in training

arXiv:2605.13829v1 Announce Type: cross Abstract: We introduce Negation Neglect, where finetuning LLMs on documents that flag a claim as false makes them believe the claim is true. For example, models

Industry

Vaporware or not? Aptera assembles its first five validation models.

Aptera Motors announced that five validation vehicles have been driven off its newly established low-volume validation assembly line in Carlsbad, California in May 2026. Running multiple vehicles thro

industryars-technica

AcuityBench: Evaluating Clinical Acuity Identification and Uncertainty Alignment

arXiv:2605.11398v1 Announce Type: cross Abstract: We introduce AcuityBench, a benchmark for evaluating whether language models identify the appropriate urgency of care from user medical presentations.

AutoMonitor-Bench: Evaluating the Reliability of LLM-Based Misbehavior Monitor

arXiv:2601.05752v3 Announce Type: replace Abstract: We introduce AutoMonitor-Bench, the first benchmark designed to systematically evaluate the reliability of LLM-based misbehavior monitors across div

Crash Assessment via Mesh-Based Graph Neural Networks and Physics-Aware Attention

arXiv:2605.11784v1 Announce Type: cross Abstract: Full-vehicle crash simulations are computationally expensive, limiting their use in iterative design exploration. This work investigates learned hybri

Multi-Task Representation Learning for Conservative Linear Bandits

arXiv:2605.12176v1 Announce Type: new Abstract: This paper presents the Constrained Multi-Task Representation Learning (CMTRL) framework for linear bandits. We consider T linear bandit tasks in a d di

Qwen-Scope: Turning Sparse Features into Development Tools for Large Language Models

arXiv:2605.11887v1 Announce Type: new Abstract: Large language models have achieved remarkable capabilities across diverse tasks, yet their internal decision-making processes remain largely opaque, li

Robust Promptable Video Object Segmentation

arXiv:2605.12006v1 Announce Type: new Abstract: The performance of promptable video object segmentation (PVOS) models substantially degrades under input corruptions, which prevents PVOS deployment in

Urban Risk-Aware Navigation via VQA-Based Event Maps for People with Low Vision

arXiv:2605.11782v1 Announce Type: new Abstract: Visual impairment affects hundreds of millions of people worldwide, severely limiting their ability to navigate urban environments safely and independen

Aligning Agents via Planning: A Benchmark for Trajectory-Level Reward Modeling

arXiv:2604.08178v2 Announce Type: replace Abstract: In classical Reinforcement Learning from Human Feedback (RLHF), Reward Models (RMs) serve as the fundamental signal provider for model alignment. As

ChatbotManip: A Dataset to Facilitate Evaluation and Oversight of Manipulative Chatbot Behaviour

arXiv:2506.12090v2 Announce Type: replace Abstract: This paper introduces ChatbotManip, a novel dataset for studying manipulation in Chatbots. It contains simulated generated conversations between a c

Decomposing and Steering Functional Metacognition in Large Language Models

arXiv:2605.08942v1 Announce Type: new Abstract: Large language models (LLMs) increasingly exhibit behaviors suggesting awareness of their evaluation context, often adapting their reasoning strategies

GONE: Structural Knowledge Unlearning via Neighborhood-Expanded Distribution Shaping

arXiv:2603.12275v1 Announce Type: cross Abstract: Unlearning knowledge is a pressing and challenging task in Large Language Models (LLMs) because of their unprecedented capability to memorize and dige

GUARD: Guideline Upholding Test through Adaptive Role-play and Jailbreak Diagnostics for LLMs

arXiv:2508.20325v3 Announce Type: replace-cross Abstract: As Large Language Models (LLMs) become increasingly integral to various domains, their potential to generate harmful responses has prompted si

HiDrive: A Closed-Loop Benchmark for High-Level Autonomous Driving

arXiv:2605.09972v1 Announce Type: cross Abstract: End-to-end autonomous driving has witnessed rapid progress, yet existing benchmarks are increasingly saturated, with state-of-the-art models achieving

Local Ai

Interactive Critique-Revision Training for Reliable Structured LLM Generation

arXiv:2605.08327v1 Announce Type: cross Abstract: In structured decision-making workflows such as form filling, compliance checking, and maintenance reporting, LLM outputs must be locally correct, glo

local-aiarxiv-cs-ai

Less Diverse, Less Safe: The Indirect But Pervasive Risk of Test-Time Scaling in Large Language Models

arXiv:2510.08592v3 Announce Type: replace-cross Abstract: Test-Time Scaling (TTS) improves LLM reasoning by exploring multiple candidate responses and then operating over this set to find the best out

Measuring What Matters: Benchmarking Generative, Multimodal, and Agentic AI in Healthcare

arXiv:2605.08445v1 Announce Type: new Abstract: AI models are increasingly deployed in live clinical environments where they must perform reliably across complex, high-stakes workflows that standard t

Med-StepBench: A Hierarchical Reasoning Framework for Evaluating Hallucinations in Medical Vision-Language Models

arXiv:2605.10002v1 Announce Type: new Abstract: Large vision-language models (VLMs) demonstrate strong performance in medical image understanding, but frequently generate clinically plausible yet inco

Medical Incident Causal Factors and Preventive Measures Generation Using Tag-based Example Selection in Few-shot Learning

arXiv:2605.10025v1 Announce Type: cross Abstract: In high-stakes domains such as healthcare, the reliability of Large Language Models (LLMs) is critical, particularly when generating clinical insights

Navigating the Sea of LLM Evaluation: Investigating Bias in Toxicity Benchmarks

arXiv:2605.10639v1 Announce Type: new Abstract: The rapid adoption of LLMs in both research and industry highlights the challenges of deploying them safely and reveals a gap in the systematic evaluati

Phoenix-VL 1.5 Medium Technical Report

arXiv:2605.10391v1 Announce Type: cross Abstract: We introduce Phoenix-VL 1.5 Medium, a 123B-parameter natively multimodal and multilingual foundation model, adapted to regional languages and the Sing

Local Ai

Playing games with knowledge: AI-Induced delusions need game theoretic interventions

arXiv:2605.08409v1 Announce Type: new Abstract: Conversational AI has a fundamental flaw as a knowledge interface: sycophantic chatbots induce epistemic entrenchment and delusional belief spirals even

local-aiarxiv-cs-ai

Position: AI Security Policy Should Target Systems, Not Models

arXiv:2605.09504v1 Announce Type: cross Abstract: We present swarm-attack, an open-source adversarial testing framework in which multiple lightweight LLM agents coordinate through shared memory, paral

Local Ai

PYTHALAB-MERA: Validation-Grounded Memory, Retrieval, and Acceptance Control for Frozen-LLM Coding Agents

arXiv:2605.08468v1 Announce Type: cross Abstract: Local LLM-based coding agents increasingly work in settings where correctness is earned through execution feedback, persistent state, and bounded repa

local-aiarxiv-cs-ai

Scaling Vision Models Does Not Consistently Improve Localisation-Based Explanation Quality

arXiv:2605.10142v1 Announce Type: cross Abstract: Artificial intelligence models are increasingly scaled to improve predictive accuracy, yet it remains unclear whether scale improves the quality of po

Scam2Prompt: A Scalable Framework for Auditing Malicious Scam Endpoints in Production LLMs

arXiv:2509.02372v3 Announce Type: replace-cross Abstract: Large Language Models have become critical to modern software development, but their reliance on uncurated web-scale datasets for training int

Steerable but Not Decodable: Function Vectors Operate Beyond the Logit Lens

arXiv:2604.02608v2 Announce Type: replace Abstract: Activation steering presupposes that task-relevant behaviors correspond to linear directions in activation space -- directions that should both stee

The Echo Amplifies the Knowledge: Somatic Marker Analogues in Language Models via Emotion Vector Re-Injection

arXiv:2605.08611v1 Announce Type: new Abstract: Current language model memory systems store what happened but not how it felt. This distinction -- between semantic memory (knowing about a past event)

The Extrapolation Cliff in On-Policy Distillation of Near-Deterministic Structured Outputs

arXiv:2605.08737v1 Announce Type: cross Abstract: On-policy distillation (OPD) is widely used for LLM post-training. When pushed with a reward-extrapolation coefficient lambda > 1, the student can lif

What's the plan? Metrics for implicit planning in LLMs and their application to rhyme generation and question answering

arXiv:2601.20164v2 Announce Type: replace-cross Abstract: Prior work suggests that language models, while trained on next token prediction, show implicit planning behavior: they may select the next to