safety

Amplifying, Not Learning: Fine-Tuned AI Text Detectors Amplify a Pretrained Direction

arXiv:2605.21653v1 Announce Type: cross Abstract: AI text detectors amplify a pretrained typicality axis; they do not construct an AI-vs-human boundary. On raw encoders before any task supervision, pr

Safety

Beyond Pixels: Learning Invariant Rewards for Real-World Robotics From a Few Demonstrations

DGX agent

arXiv:2605.22123v1 Announce Type: new Abstract: Designing reward functions that generalize beyond controlled laboratory settings remains a fundamental challenge in reinforcement learning for robotics.

safetyarxiv-cs-ro

22 May 2026

Bounding-Box Trajectories Matter for Video Anomaly Detection

arXiv:2605.21957v1 Announce Type: new Abstract: Video anomaly detection is critical for public safety and security, yet remains highly challenging despite extensive research due to large variations in

Branch-Stochastic Model Predictive Control for Motion Planning under Multi-Modal Uncertainty with Scenario Clustering

arXiv:2605.22600v1 Announce Type: new Abstract: Motion planning for autonomous driving must account for multi-modal uncertainty in both the intentions and trajectories of surrounding vehicles. Handlin

Case-Aware Medical Image Classification with Multimodal Knowledge Graphs and Reliability-Guided Refinement

arXiv:2605.22547v1 Announce Type: new Abstract: Deep learning has brought significant progress to medical image classification, yet most existing methods still rely on isolated visual evidence and can

Circle-RoPE: Cone-like Decoupled Rotary Positional Embedding for Large Vision-Language Models

arXiv:2505.16416v3 Announce Type: replace Abstract: Rotary Position Embedding (RoPE) is widely adopted in large language models, but when applied to vision-language models (VLMs) it couples text and i

Closed-Loop Sim-to-Real Reinforcement Learning for Deformable Microfiber Shape Control

arXiv:2605.21688v1 Announce Type: new Abstract: Autonomous contact-based micromanipulation is challenging because surface and interfacial interactions at the microscale are difficult to model accurate

CR4T: Rewrite-Based Guardrails for Adolescent LLM Safety

arXiv:2605.21609v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly embedded in adolescent digital environments, mediating information seeking, advice, and emotionally sensit

Depth Augmented and FE Free 3D/2D Liver Registration for Laparoscopic Liver AR

arXiv:2602.17517v2 Announce Type: replace Abstract: Augmented reality (AR) guidance in laparoscopic liver surgery requires accurate registration of preoperative 3D models to intraoperative 2D video, b

Diagnosis Is Not Prescription: Linguistic Co-Adaptation Explains Patching Hazards in LLM Pipelines

arXiv:2605.21958v1 Announce Type: new Abstract: When a multi-module LLM agent fails, the module most responsible for the failure is not necessarily the best place to intervene. We demonstrate this Dia

Discovering Implicit Large Language Model Alignment Objectives

arXiv:2602.15338v2 Announce Type: replace-cross Abstract: Large language model (LLM) alignment relies on complex reward signals that often obscure the specific behaviors being incentivized, creating c

Distributed Image Compression with Multimodal Side Information at Extremely Low Bitrates

arXiv:2605.22061v1 Announce Type: new Abstract: Distributed Image Compression (DIC) is crucial for multi-view transmission, especially when operating at extremely low bitrates (< 0.1 bpp). Its core ch

Does seem that way. Happy IPO season!

Gary Marcus comments positively on what appears to be a strong period for initial public offerings (IPOs) in a brief social media post. The post suggests optimism about market conditions favorable to

Echo4DIR: 4D Implicit Heart Reconstruction from 2D Echocardiography Videos

arXiv:2605.22066v1 Announce Type: new Abstract: Reconstructing 4D (3D+t) cardiac geometry from sparse 2D echocardiography is highly desirable yet fundamentally challenged by geometric ambiguity and te

Eight key points from the most recent essay in the “AI as Normal Technology” series by @sayashk and me. Do AI Risks Require Extraordinary Go…

Eight key points from the most recent essay in the “AI as Normal Technology” series by @sayashk and me. Do AI Risks Require Extraordinary Government Intervention? 1. There is general consensus that AI

safetyyann-lecun--x

Energy-Gated Attention: Spectral Salience as an Inductive Bias for Transformer Attention

arXiv:2605.21842v1 Announce Type: cross Abstract: Standard transformer attention computes pairwise similarity between queries and keys, treating all tokens as equally salient regardless of their intri

Enhancing Multimodal Large Language Models for Safety-Critical Driving Video Analysis

arXiv:2605.22185v1 Announce Type: new Abstract: Recent advancements in Multimodal Large Language Models (MLLMs) have demonstrated impressive capabilities in general visual understanding. However, thei

FlyRoute: Self-Evolving Agent Profiling via Data Flywheel for Adaptive Task Routing

arXiv:2605.22057v1 Announce Type: new Abstract: Enterprise routers assign queries to expert agents, yet deployed profiles stay static while agents evolve (prompts, tools, models), and developers rarel

Focusing Where Vision Matters: Selective Training for Large Vision Language Models via Visual Information Gain

arXiv:2602.17186v2 Announce Type: replace Abstract: Large Vision Language Models (LVLMs) have achieved remarkable progress, yet they often suffer from language bias, producing answers without relying

From Abstraction to Instantiation: Learning Behavioral Representation for Vision-Language-Action Model

arXiv:2605.22671v1 Announce Type: new Abstract: Vision-Language-Action (VLA) models often suffer from performance degradation under distribution shifts, as they struggle to learn generalized behavior

Fuck this OpenAI employee, seriously fuck him. Also read this quantitative study, which I had nothing to do with, that says my technical pre…

Fuck this OpenAI employee, seriously fuck him. Also read this quantitative study, which I had nothing to do with, that says my technical predictions have bee largely correct. https://github.com/davego

“Gary Marcus, former CEO of Geometric Intelligence and New York University professor, joins @SquawkStreet @CNBC to discuss why he still beli…

“Gary Marcus, former CEO of Geometric Intelligence and New York University professor, joins @SquawkStreet @CNBC to discuss why he still believes OpenAI could be the ‘WeWork’ of AI, why he finds Anthro

GenEvolve: Self-Evolving Image Generation Agents via Tool-Orchestrated Visual Experience Distillation

arXiv:2605.21605v1 Announce Type: new Abstract: Open-ended image generation is no longer a simple prompt-to-image problem. High-quality generation often requires an agent to combine a model's internal

GLeVE: Graph-Guided Lesion Grounding with Proposal Verification in 3D CT

arXiv:2605.22619v1 Announce Type: new Abstract: Grounding radiology report descriptions to 3D CT volumes is essential for verifiable clinical interpretation, yet remains challenging due to the semanti

Google files its appeal of the US federal ruling deeming it an illegal search monopolist, arguing it 'prevailed in the marketplace fair and square' (Lauren Feiner/The Verge)

Lauren Feiner / The Verge: Google files its appeal of the US federal ruling deeming it an illegal search monopolist, arguing it “prevailed in the marketplace fair and square” — It wants to throw out t

safetytechmeme

Governance by Construction for Generalist Agents

arXiv:2605.20874v1 Announce Type: new Abstract: Enterprise agents are increasingly expected to operate autonomously across tools and interfaces, yet production deployments require governance by constr

Governance by Design: Architecting Agentic AI for Organizational Learning and Scalable Autonomy

arXiv:2605.20210v1 Announce Type: cross Abstract: Agentic AI systems - systems that can pursue goals through multi-step planning and tool-mediated action with limited direct supervision - are moving f

Harder to Defend: Towards Chinese Toxicity Attacks via Implicit Enhancement and Obfuscation Rewriting

arXiv:2605.22258v1 Announce Type: new Abstract: Large language models (LLMs) require robust toxicity evaluation beyond explicit wording. This setting remains underexplored in Chinese, where toxicity m

Hierarchical Variational Policies for Reward-Guided Diffusion

arXiv:2605.21661v1 Announce Type: cross Abstract: Adapting pretrained diffusion models to downstream objectives such as inverse problems often requires expensive test-time guidance or optimization. We

How can reasoning capability empower the AI copilot robot in endoscopic surgery

arXiv:2605.22322v1 Announce Type: new Abstract: Reasoning capability has significantly advanced complex logical inference and robotic decision-making in general domains. However, its potential in the

I just ran into a guy I know in NYC who said he’s doing a temporary gig for OpenAI. According to him, OpenAI is paying hundreds of people an…

I just ran into a guy I know in NYC who said he’s doing a temporary gig for OpenAI. According to him, OpenAI is paying hundreds of people and families in New York to install 360-degree cameras through

i suspect that this is a wild underestimate, both ignoring the costs in developing the model and ignoring the fact that many questions may w…

i suspect that this is a wild underestimate, both ignoring the costs in developing the model and ignoring the fact that many questions may well have been posed that didn’t succeed. If this is true, us

If Elon Musk was really serious about his proposal for 'universal high income,' @scottsantens tells me, he could start funding it right now.…

If Elon Musk was really serious about his proposal for 'universal high income,' @scottsantens tells me, he could start funding it right now. But he isn't doing that. Instead, he's railing about the ve

LACO: Adaptive Latent Communication for Collaborative Driving

arXiv:2605.22504v1 Announce Type: cross Abstract: Collaborative driving aims to improve safety and efficiency by enabling connected vehicles to coordinate under partial observability. Recent approache

LANG: Reinforcement Learning for Multilingual Reasoning with Language-Adaptive Hint Guidance

arXiv:2605.22567v1 Announce Type: new Abstract: Reinforcement learning has proven effective for enhancing multi-step reasoning in large language models (LLMs), yet its benefits have not fully translat

Learning Altruistic Collaboration in Heterogeneous Multi-Team Systems

arXiv:2605.21723v1 Announce Type: new Abstract: This paper studies heterogeneous multi-team collaboration through dynamic robot allocation, where robots are treated as transferable resources. Leveragi

Learning to Configure Agentic AI Systems

arXiv:2602.11574v3 Announce Type: replace Abstract: Configuring LLM-based agent systems involves choosing workflows, tools, token budgets, and prompts from a large combinatorial design space, and is t

Learning to Evolve: Multi-modal Interactive Fields for Robust Humanoid Navigation in Dynamic Environments

arXiv:2605.21935v1 Announce Type: new Abstract: Safe manipulation-oriented navigation for humanoid robots requires scene memory that remains reliable under locomotion-induced perceptual distortion, en

Making the Discrete Continuous: Synthetic RAW Augmentations for Fine-Grained Evaluation of Person Detection Performance in Low Light

arXiv:2605.22455v1 Announce Type: new Abstract: Real-world deployment of AI vision models is both fueled and limited by the data available for training and testing. Real datasets are sparse and uneven

Mapping Tomato Cropping Systems in California Using AlphaEarth Geospatial Embeddings and Deep Learning Analysis

arXiv:2605.21804v1 Announce Type: cross Abstract: Field-scale crop maps support supply-chain forecasting and policy, yet statewide crop identification still often depends on retrospective surveys or r

Modeling Emotional Dynamics in Agent-to-Agent Interactions on Moltbook

arXiv:2605.20442v1 Announce Type: cross Abstract: Generative AI systems are increasingly deployed as interactive agents in online environments, such as a social network called Moltbook. In Moltbook, l

Modeling Pathology-Like Behavioral Patterns in Language Models Through Behavioral Fine-Tuning

arXiv:2605.22356v1 Announce Type: new Abstract: Large language models are increasingly used as computational tools for modeling human-like behavior. We introduce a behavioral induction framework that

Moral Semantics Survive Machine Translation: Cross-Lingual Evidence from Moral Foundations Corpora

arXiv:2605.22660v1 Announce Type: new Abstract: Moral language is subtle and culturally variable, making it difficult to translate faithfully across languages. Idiomatic expressions, slang, and cultur

my private index of desperation is how many people lie about me. we are in yellow alert territory

Gary Marcus expresses frustration about false claims being made about him, using a 'desperation index' metric based on the volume of lies circulated in his name. He indicates the situation has reached

NaviAgent: Graph-Driven Bilevel Planning for Scalable Tool Orchestration

arXiv:2506.19500v3 Announce Type: replace-cross Abstract: Large Language Models (LLMs) increasingly act as function-call agents that invoke external tools to tackle tasks beyond their static knowledge

Noise-Space Attribution and Control of Chunk-Boundary Artifact

arXiv:2603.11642v2 Announce Type: replace Abstract: Action chunking is widely used in generative visuomotor policies, yet the recurring execution discontinuities at chunk boundaries still lack a mecha

Non-Contact Vibration-Based Damage Detection of Civil Structures Using a Cost-Effective Autonomous UAV

arXiv:2605.21914v1 Announce Type: new Abstract: This paper presents a non-contact approach for vibration-based structural damage detection using an autonomous and customized cost-effective unmanned ae

nope you are. because openai will sell your most private data to the government.

This post appears to express concerns about OpenAI's data privacy practices and potential government data sharing, though the claim lacks specific evidence or context. Gary Marcus, an AI researcher an