Search: “safety”

arXiv:2508.17916v2 Announce Type: replace Abstract: Depth estimation is a foundational component for 3D reconstruction in minimally invasive endoscopic surgeries. However, existing monocular depth est

local-aiarxiv-cs-cv

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Model Releases

FlameVQA: A Physically-Grounded UAV Wildfire VQA Benchmark with Radiometric Thermal Supervision

DGX agent

arXiv:2606.27128v1 Announce Type: new Abstract: Wildfire monitoring from UAVs requires reliable reasoning over complex aerial scenes, where smoke, scale variation, and occlusions often limit RGB-only

model-releasesarxiv-cs-cv

26 Jun 2026

Jailbreaking for the Average Jane: Choosing Optimal Jailbreaks via Bandit Algorithms for Automatically Enhanced Queries

arXiv:2606.26936v1 Announce Type: cross Abstract: With a profusion of jailbreaks for LLMs now widely known, a growing concern is that non-expert malicious actors ('the average Jane') could elicit acti

Reproducibility Study of 'AlphaEdit: Null-Space Constrained Knowledge Editing for Language Models'

arXiv:2606.26783v1 Announce Type: cross Abstract: Fang et al. (2025) introduced a null-space constrained projection, named AlphaEdit, for locate-then-edit knowledge editing methods, theoretically guar

When Actions Go Off-Task: Detecting and Correcting Misaligned Actions in Computer-Use Agents

arXiv:2602.08995v2 Announce Type: replace Abstract: Computer-use agents (CUAs) have made tremendous progress in the past year, yet they still frequently produce misaligned actions that deviate from th

How Reliable Is Your Jailbreak Judge? Calibration and Adversarial Robustness of Automated ASR Scoring

arXiv:2606.25487v1 Announce Type: new Abstract: Almost every paper on LLM jailbreaks and prompt injection reports an attack-success rate (ASR), and that number is assigned not by people but by an auto

SpeechEQ: Benchmarking Emotional Intelligence Quotient in Socially Aware Voice Conversational Models

arXiv:2606.25990v1 Announce Type: new Abstract: As multimodal conversational systems increasingly engage in spoken interaction, their ability to navigate paralinguistic social cues has become a critic

The 4/elta Bound: Designing Predictable LLM-Verifier Systems for Formal Method Guarantee

model-releasesarxiv-cs-lg

arXiv:2512.02080v3 Announce Type: replace-cross Abstract: The integration of Formal Verification tools with Large Language Models (LLMs) offers a path to scale software verification beyond manual work

Toward Low-Latency Vision-Language Models with Doubly-Correct Predictions in Egocentric Visual Understanding

arXiv:2606.25160v1 Announce Type: cross Abstract: The rapid rise of Vision-Language Models (VLMs) in egocentric visual understanding has made low-latency inference in human-robot collaborative (HRC) t

Uncertainty Quantification for Computer-Use Agents: A Benchmark across Vision-Language Models and GUI Grounding Datasets

arXiv:2606.25760v1 Announce Type: cross Abstract: Computer-use agents turn vision-language model (VLM) predictions into executable GUI clicks, so reliable uncertainty estimates are essential for rejec

What Intermediate Layers Know: Detecting Jailbreaks from Entropy Dynamics

arXiv:2606.25182v1 Announce Type: new Abstract: Jailbreak attacks reveal a persistent weakness in aligned Large Language Models: carefully crafted prompts can elicit policy-violating responses despite

WOLF-VLA: Whole-Body Humanoid Optimal Locomotion Framework for Vision-Language-Action Learning

arXiv:2606.25591v1 Announce Type: new Abstract: Vision-Language-Action (VLA) models have recently demonstrated strong generalization in robotic manipulation, yet their applicability to whole-body, con

Industry

Elon Musk denies Tesla’s Autopilot caused crash that killed grandmother

A Tesla Model 3 crashed into a residential home in Katy, Texas, killing a 76-year-old grandmother, with the driver claiming Autopilot was active at the time. Elon Musk disputed the claim, and Tesla's

industryars-technica

ESBMC-PLC+: A Unified IEC~61131-3 Formal Verification Framework as a PLCverif Successor

arXiv:2606.23870v1 Announce Type: cross Abstract: PLCverif is the most mature open-source platform for PLC formal verification, developed at CERN and in production use since 2019. Yet it has two funda

How Much Can We Trust LLM Search Agents? Measuring Endorsement Vulnerability to Web Content Manipulation

arXiv:2606.16821v2 Announce Type: replace Abstract: Large language model (LLM)-based search agents synthesize open-web content into actionable recommendations on behalf of users, creating a risk that

Neuro-Symbolic Drive: Rule-Grounded Faithful Reasoning for Driving VLAs

arXiv:2606.23938v1 Announce Type: new Abstract: Driving VLA models incorporating Chain-of-Thought (CoT) reasoning are attractive because they leverage pretrained VLM representations and expose interme

Poster: Exploring the Limits of Audio-Based Detection of Turkish Phone Call Scams

arXiv:2606.24523v1 Announce Type: cross Abstract: Scam phone calls exploit vulnerable communities worldwide, yet research on detection has focused almost exclusively on English and other high-resource

RASC+: Retrieval-Constrained LLM Adjudication for Clinical Value Set Authoring

arXiv:2606.23992v1 Announce Type: cross Abstract: Clinical value sets define the standardized terminology codes used in quality measurement, phenotyping, cohort construction, and clinical decision sup

T2D-Bench: Evidence-Gated Evaluation of LLM Outputs for Type 2 Diabetes Using a Multi-Layer Clinical-Lifestyle Knowledge Graph

arXiv:2606.24145v1 Announce Type: new Abstract: Large language models (LLMs) can produce clinically fluent recommendations for type 2 diabetes while failing to satisfy guideline constraints or explici

Research

The best way to understand a complex system is via edge cases and failure modes, because they define the contour of the system.

researchfrancois-chollet--x

Edge cases and failure modes reveal the fundamental boundaries and constraints of complex systems more effectively than typical operations, making them valuable for understanding system behavior and l

UniDrive: A Unified Vision-Language and Grounding Framework for Interpretable Risk Understanding in Autonomous Driving

arXiv:2606.24759v1 Announce Type: cross Abstract: Recent multimodal large language models (MLLMs) have shown strong potential for autonomous driving scene understanding, yet existing methods still fac

Attacking the Trusted Imagination: Oracle-Level Integrity Attacks on Imagine-then-Act World Models

model-releasesarxiv-cs-lg

arXiv:2606.22966v1 Announce Type: new Abstract: Many recent vision-language-action (VLA) policies adopt an imagine-then-act design. A world-action model (WAM) first imagines a short future as a latent

Hierarchical Adversarial Bandits for Online Configuration Optimization

model-releasesarxiv-cs-lg

arXiv:2505.19061v2 Announce Type: replace Abstract: Motivated by Online Configuration Optimization in large, dynamic parameter spaces, this work studies the nonstochastic multi-armed bandit (MAB) prob

Large Language Model-Assisted Cleaning of Report-Derived Labels in a Large-Scale Chest CT Dataset

arXiv:2606.22382v1 Announce Type: cross Abstract: Purpose: To evaluate whether large language model (LLM)-assisted label cleaning can identify label-report discordance in CT-RATE, a large-scale public

Local Ai

Mind the Noise: Sensitivity of Transformer-based Interaction-Aware Trajectory Prediction Models to Noisy Data

arXiv:2606.21344v1 Announce Type: cross Abstract: Trajectory prediction allows autonomous vehicles to anticipate the future behavior of surrounding objects (or agents) and, accordingly, maximize the s

local-aiarxiv-cs-lg

Open models, global networks: How AT&T and GSMA are accelerating telecom innovation with Gemma

model-releasesgoogle-cloud-ai

Telecommunications is an incredibly complex, highly specialized domain. Modern mobile networks are inherently multi-vendor, featuring diverse and often proprietary data structures. While AI has made m

R2HandoverSim: A Simulation Framework and Benchmark for Robot-to-Human Object Handovers

arXiv:2606.21011v1 Announce Type: new Abstract: We present R2HandoverSim, a simulation benchmark for robot-to-human (R2H) object handovers. Although R2H handover methods have advanced rapidly, the lac

SparseWorld: Enhancing End-to-End Autonomous Driving via World Models with Sparse Scene Representation

arXiv:2605.24354v2 Announce Type: replace Abstract: Recently, world models have made significant progress in enhancing end-to-end driving systems through both future situation forecasting and improved

Zero-Shot Vision-Language Models for Classroom Engagement Recognition: A Benchmark Study of Prompt Sensitivity and Cross-Dataset Generalization

arXiv:2606.21861v1 Announce Type: new Abstract: Automated classroom engagement recognition holds substantial promise for scalable learning analytics, yet the suitability of modern Vision-Language Mode

Applications

1,250 hp hybrid Corvette shatters the Pikes Peak production record

Chevrolet's Corvette ZR1X, powered by a hybrid all-wheel drive system with 1,250 horsepower, set a new Pikes Peak production car record at the 104th running of the hill climb. Driven by IndyCar vetera

applicationsars-technica

Daybreak: Tools for securing every organization in the world

Daybreak is an OpenAI initiative focused on developing and distributing security tools designed to protect organizations globally from cyber threats. The program aims to democratize access to advanced

model-releasesopenai

Industry

GM installs robots at flagship EV factory after laying off 1,300 workers

General Motors temporarily laid off 1,300 workers at its Factory Zero EV plant due to sluggish demand for battery-powered vehicles , following the discontinuation of the $7,500 federal EV tax credit .

industryars-technica

Industry

How Anthropic may have talked itself into an AI export ban

The US government ordered Anthropic to cut off foreign access to its Mythos 5 and Claude Fable 5 AI models in June 2026, citing national security concerns related to jailbreak vulnerabilities that cou

industryars-technica

Tools

Red-Teaming after Mythos — Zico Kolter & Matt Fredrikson, Gray Swan

This episode likely discusses red-teaming methodologies and adversarial testing approaches in AI systems, featuring security researchers Zico Kolter and Matt Fredrikson from Carnegie Mellon University

toolslatent-space

Agent Skill Evaluation and Evolution: Frameworks and Benchmarks

arXiv:2606.11435v1 Announce Type: new Abstract: The growth of agent skills has transformed how agentic systems are built, evaluated, and deployed. As skill libraries continue to scale, rigorous evalua

11 Jun 2026

Making Models Unmergeable via Scaling-Sensitive Loss Landscape

arXiv:2601.21898v2 Announce Type: replace Abstract: The rise of model hubs has made it easier to access reusable model components, making model merging a practical tool for combining capabilities. Yet

11 Jun 2026

LLM-Based Code Documentation Generation and Multi-Judge Evaluation

arXiv:2606.09852v1 Announce Type: cross Abstract: High-quality source code documentation is vital yet often neglected, especially in critical domains like healthcare where reliability and maintainabil

10 Jun 2026

Uncertainty-Aware Motion Planning for Autonomous Driving in Mixed Traffic Environment

arXiv:2606.09958v1 Announce Type: cross Abstract: In mixed-traffic environments where autonomous and human-driven vehicles may co-exist, motion planning for autonomous vehicles requires anticipating t

10 Jun 2026

Local Ai

AI-Native Closed-Loop Security for 6G-Enabled Cyber-Physical Systems: From Edge Detection to Network-Wide Mitigation

arXiv:2606.08173v1 Announce Type: cross Abstract: In sixth-generation (6G) networks, billions of cyber-physical systems (CPSs) - autonomous vehicles, smart grids, industrial robots, and remote-surgica

local-aiarxiv-cs-lg

Anthropic says Fable 5 has invisible safeguards that use prompt modification, steering vectors, or PEFT to limit its effectiveness for building frontier LLMs (Matthias Bastian/The Decoder)

Matthias Bastian / The Decoder: Anthropic says Fable 5 has invisible safeguards that use prompt modification, steering vectors, or PEFT to limit its effectiveness for building frontier LLMs — Key Poin

model-releasestechmeme

Beyond Goodhart's Law: A Dynamic Benchmark for Evaluating Compliance in Multi-Agent Systems

arXiv:2606.07805v1 Announce Type: new Abstract: The rapid evolution of Large Language Models (LLMs) from passive assistants to autonomous, execution-capable agents has introduced critical operational

Explaining Black-Box Language Models: Learning to Optimize Linguistically-Structured Word Subsets

arXiv:2606.08497v1 Announce Type: new Abstract: As deep language models (DLMs) are increasingly deployed in high-stakes domains such as healthcare, understanding their decision rationale becomes param

Initial impressions of Claude Fable 5

model-releasessimon-willison

I didn't have early access to today's Claude Fable 5 release, but I've spent the past ~5.5 hours putting it through its paces. My initial impressions are that this is something of a beast. It's slow,

Learning from Human Driving: A Human-in-the-Loop Online Behavior Cloning Framework for Autonomous Driving

arXiv:2606.08170v1 Announce Type: new Abstract: With the evolution of large foundation models (LFMs), data-driven autonomous driving has made significant strides. However, existing paradigms still fac

PLAGUE: Plug-and-play framework for Lifelong Adaptive Generation of Multi-turn Exploits

arXiv:2510.17947v3 Announce Type: replace-cross Abstract: Large Language Models (LLMs) are improving at an exceptional rate. With the advent of agentic workflows, multi-turn dialogue has become the de

ProbeAct: Probe-Guided Training-Free Failure Recovery in Vision-Language-Action Models

arXiv:2606.09740v1 Announce Type: new Abstract: Vision-Language-Action (VLA) models demonstrate strong perfor-1 mance on language-conditioned robotic manipulation within their training dis-2 tribution

RecurGuard: Runtime Monitoring for Reasoning-Token Consumption Attacks

arXiv:2606.07968v1 Announce Type: cross Abstract: Reasoning-capable large language models can be induced to spend their generation budget on injected decoy tasks rather than answering the user's quest

RiskNet: A large-scale dataset of AI risk incidents from news with alignment and multi-dimensional annotations

arXiv:2606.08376v1 Announce Type: cross Abstract: As artificial intelligence (AI) systems are increasingly deployed across socially consequential domains, reports of AI-related harms and failures have