safety

Pope Leo warns AI revolution driven by ‘idolatry of profit’ https://ft.trib.al/UGP9H8b

Pope Leo has warned that the artificial intelligence revolution is being driven primarily by profit motives rather than ethical considerations, characterizing this as a form of 'idolatry of profit.' T

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Safety

Precise: SDE-Consistent Stochastic Sampling for RL Post-Training of Flow-Matching Models

DGX agent

arXiv:2605.23522v1 Announce Type: cross Abstract: Reinforcement learning (RL) has become an effective way to improve prompt alignment and perceptual quality in diffusion and flow-matching generators.

safetyarxiv-cs-ai

25 May 2026

Prudent-Banker: No Extra Fees for Baseline Safety in Adversarial Bandits With and Without Delays

arXiv:2605.23351v1 Announce Type: new Abstract: We study adversarial multi-armed bandits with and without delayed feedback under a safety-aware goal: achieving minimax-optimal worst-case regret while

Pure Exploration for a Good Policy in Reinforcement Learning with Bandit Feedback

arXiv:2605.23182v1 Announce Type: new Abstract: Pure exploration in episodic Reinforcement Learning has primarily focused on Best Policy Identification (BPI), which seeks to identify a (near)-optimal

RAG-Pull: Turning Retrieval into a Code-Injection Channel via Invisible Unicode Perturbations

arXiv:2510.11195v2 Announce Type: replace-cross Abstract: Retrieval-Augmented Generation (RAG) increases the reliability and trustworthiness of the LLM response and reduces hallucination by eliminatin

RAG4Outcome: A Retrieval-Augmented Multimodal Framework for Prognostic Prediction in Chronic Osteomyelitis

arXiv:2605.22833v1 Announce Type: cross Abstract: Chronic osteomyelitis presents substantial prognostic challenges due to its high recurrence risk and complex postoperative recovery trajectories. Trad

ReCoVer: Resilient LLM Pre-Training System via Fault-Tolerant Collective and Versatile Workload

arXiv:2605.11215v2 Announce Type: replace-cross Abstract: Pre-training large language models on massive GPU clusters has made hardware faults routine rather than rare, driving the need for resilient t

Reflex: Reinforcement Learning with Reflection Symmetry Exploitation in State-Based Continuous Control

arXiv:2605.23415v1 Announce Type: cross Abstract: Reinforcement learning has long struggled with poor sample efficiency. One promising approach to mitigate this problem is leveraging group-invariant M

Reinforcement Learning with Verifiable yet Noisy Rewards under Imperfect Verifiers

arXiv:2510.00915v4 Announce Type: replace-cross Abstract: Reinforcement Learning with Verifiable Rewards (RLVR) replaces costly human labeling with automated verifiers. To reduce verifier hacking, man

Relevant Walk Search for Explaining Graph Neural Networks

arXiv:2605.23673v1 Announce Type: new Abstract: Graph Neural Networks (GNNs) have become important machine learning tools for graph analysis, and its explainability is crucial for safety, fairness, an

Remote Teleoperation of Endovascular Intervention Robots: A Systematic Review

arXiv:2605.22889v1 Announce Type: new Abstract: Remote robotic-assisted endovascular intervention offers a promising approach to reduce clinician radiation exposure and physical strain, while extendin

Robotic Strawberry Harvesting with Robust Vision and Deep Reinforcement Learning based Sim-to-Real Control

arXiv:2605.23863v1 Announce Type: new Abstract: This study presents a closed-loop robotic strawberry harvesting system that combines a robust vision module, simulation-trained deep reinforcement learn

Safe Reinforcement Learning with Preference-based Constraint Inference

arXiv:2603.23565v2 Announce Type: replace-cross Abstract: Safe reinforcement learning (RL) is a standard paradigm for safety-critical decision making. However, real-world safety constraints can be com

SafeHarbor: Hierarchical Memory-Augmented Guardrail for LLM Agent Safety

arXiv:2605.05704v2 Announce Type: replace-cross Abstract: Recent advances in foundation models have transformed LLMs from passive conversational systems into autonomous agents capable of reasoning and

Sample-wise Targeted Adversarial Attacks on Test-time Adaptation

arXiv:2605.23411v1 Announce Type: cross Abstract: Test-time adaptation (TTA) effectively counters distribution shifts but exposes models to adversarial manipulation via the unlabeled test stream. Exis

Score-Based One-step MeanFlow Policy Optimization

arXiv:2605.23365v1 Announce Type: cross Abstract: Diffusion and flow matching have emerged as expressive policy classes in reinforcement learning, but their reliance on multi-step denoising imposes su

SCRIPT: Scalable Diffusion Policy with Multi-stage Training for Language-driven Physics-Based Humanoid Control

arXiv:2605.22894v1 Announce Type: cross Abstract: Controlling physics-based humanoids from natural-language instructions is a critical step toward general-purpose embodied agents. However, existing me

SeedER: Seed-and-Expand Retrieval from Knowledge Graphs

arXiv:2605.23753v1 Announce Type: new Abstract: Knowledge graphs (KGs) offer a rich representation for relational knowledge, but their irregular structure makes retrieval challenging: ego-graph expans

smdh

smdh Chamath Lays Out the Case for SpaceX at $2 Trillion – Starlink: the most important internet infra project since the internet itself – Rockets: underlying platform that allows everything else to h

Spatio-Temporal Similarity Volume Aggregation for Open-Vocabulary Action Recognition

arXiv:2605.23288v1 Announce Type: new Abstract: Recent Open-Vocabulary Action Recognition (OVAR) methods typically aggregate visual features into a global representation before computing text alignmen

SpinFlow: A Physics-Informed Spin Field Framework for Traffic Phase Inference and Transition Detection

arXiv:2605.23306v1 Announce Type: cross Abstract: Active traffic management (ATM) is frequently hindered by traditional macroscopic models and rigid empirical thresholds that fail to capture metastabl

sure but that shouldn’t affect the valuation of spacex as it pivots from rockets to AI at all, right? right?

Gary Marcus expresses skepticism about SpaceX's pivot from rockets to AI, questioning whether this strategic shift should impact the company's valuation. The post uses rhetorical questioning to sugges

TactileReflex: Noise-Statistics-Driven Vision-Tactile Reflex Control for Force-Sensitive Manipulation

arXiv:2605.23568v1 Announce Type: new Abstract: Manipulating fragile deformable containers, such as disposable plastic cups filled with liquid, demands real-time grip-force adaptation within an extrem

Task-Awareness Improves LLM Generations and Uncertainty

arXiv:2601.21500v2 Announce Type: replace Abstract: In many applications of LLMs, natural language responses often have an underlying structure such as representing discrete labels, numerical values,

Test-Time Training Undermines Safety Guardrails

arXiv:2605.22984v1 Announce Type: cross Abstract: Test-Time Training (TTT) is an emerging paradigm that enables models to adapt their parameters during inference, improving performance on tasks such a

The CEO of BlackRock, Larry Fink, admits that the trillions of dollars being used to build data centers and power grids will come from ordin…

The CEO of BlackRock, Larry Fink, admits that the trillions of dollars being used to build data centers and power grids will come from ordinary people’s savings accounts and pension funds, and says it

The Frenchman (@ylecun) seems incapable of giving credit where credit is due. @GaryMarcus is correct about LeCun's lack of integrity. Here h…

The Frenchman (@ylecun) seems incapable of giving credit where credit is due. @GaryMarcus is correct about LeCun's lack of integrity. Here he is rephrasing something that everyone else in the business

The FTC settles with Cox, MindSift, and 1010 Digital Works for $930K over claims they falsely said they could use phone mics to spy on users for ad targeting (Adi Robertson/The Verge)

Adi Robertson / The Verge: The FTC settles with Cox, MindSift, and 1010 Digital Works for $930K over claims they falsely said they could use phone mics to spy on users for ad targeting — More specific

safetytechmeme

The Implicit Bias of Depth: From Neural Collapse to Softmax Codes

arXiv:2605.23087v1 Announce Type: new Abstract: Neural collapse (NC) describes the structured geometry that emerges in the features and weights of trained classifiers. Recent theory suggests NC can be

The physics of AI weather models

arXiv:2605.23778v1 Announce Type: cross Abstract: Could it be that AI weather models are solving physical equations, although they may not be the equations used by conventional NWP models? We compute

The Pope rightly warns that AI must serve human dignity, not become a tool of domination or exclusion. But if we hand governments sweeping p…

The Pope rightly warns that AI must serve human dignity, not become a tool of domination or exclusion. But if we hand governments sweeping power over AI development in the name of safety, how do we pr

safetyclem-delangue--x

the tech bros think they know better than me they think they know better than @michaeljburry they think they know better than the Nobelist @…

the tech bros think they know better than me they think they know better than @michaeljburry they think they know better than the Nobelist @DAcemogluMIT they think they know better than @sundarpichai

this bit on physics is deep and actually resonates with the fact that LLMs are equally comfortable learning any world, whereas humans are bu…

this bit on physics is deep and actually resonates with the fact that LLMs are equally comfortable learning any world, whereas humans are built for our world. See also Chomsky’s 2023 conversation with

This creates problems for the @Kevinroose narrative that AGI is near.

This creates problems for the @Kevinroose narrative that AGI is near. Finally, a big name has the courage to tell it: we are nowhere near AGI. Demis Hassabis, CEO of Google DeepMind and Nobel laureate

Through the Stealth Lens: Attention-Aware Defenses Against Poisoning in RAG

arXiv:2506.04390v2 Announce Type: replace-cross Abstract: Retrieval-augmented generation (RAG) systems are vulnerable to attacks that inject poisoned passages into the retrieved context, even at low c

time to get out of index funds? they will soon be saddled with the IPOs of three giant companies that barely make a profit (or don’t make pr…

time to get out of index funds? they will soon be saddled with the IPOs of three giant companies that barely make a profit (or don’t make profits at all). I don’t personally want to be part of that. (

Transform-Invariant Generative Ray Path Sampling for Efficient Radio Propagation Modeling

arXiv:2603.01655v2 Announce Type: replace Abstract: Ray tracing has become a standard for accurate radio propagation modeling, but suffers from exponential computational complexity, as the number of c

UfM: Uncertainty from Motion for DNN Depth Estimation Using Gaussians

arXiv:2605.23098v1 Announce Type: new Abstract: Reliable uncertainty estimation is critical for deploying monocular depth deep neural networks (DNNs) in safety-critical robotic systems. Conventional u

Understanding Goal Generalisation in Sequential Reinforcement Learning

arXiv:2605.23565v1 Announce Type: cross Abstract: Reinforcement learning agents often exhibit unintended goal-directed behaviour outside their training distribution, but we currently lack a principled

UniEmo: Unifying Emotional Understanding and Generation with Learnable Expert Queries

arXiv:2507.23372v2 Announce Type: replace Abstract: Emotional understanding and generation are often treated as separate tasks, yet they are inherently complementary and can mutually enhance each othe

UniReg: A Universal Model for Controllable CT Image Registration

arXiv:2503.12868v2 Announce Type: replace Abstract: Learning-based medical image registration has matched the accuracy of conventional methods while offering superior computational efficiency. However

V-VLAPS: Value-Guided Planning for Vision-Language-Action Models

arXiv:2601.00969v2 Announce Type: replace-cross Abstract: Vision-language-action (VLA) models provide strong action priors for robotic manipulation, but their reactive behavior can fail under distribu

VI-CuRL: Stabilizing Verifier-Independent RL Reasoning via Confidence-Guided Variance Reduction

arXiv:2602.12579v2 Announce Type: replace-cross Abstract: Reinforcement Learning with Verifiable Rewards (RLVR) has emerged as a dominant paradigm for enhancing Large Language Models (LLMs) reasoning,

Vision Transformers Need Better Token Interaction

arXiv:2605.23868v1 Announce Type: new Abstract: Vision Transformers (ViTs) can learn strong image-level representations while their patch representations become less effective for dense prediction dur

What I don’t get is why so few people care that LeCun is basically a thief. I really don’t get it.

What I don’t get is why so few people care that LeCun is basically a thief. I really don’t get it. Dr. LeCun's heavily promoted Joint Embedding Predictive Architecture (JEPA, 2022) [5] is the heart of

When Determinants Are Not Enough: Private Rare Switching

arXiv:2605.23131v1 Announce Type: new Abstract: In this note, I would like to share a small research moment where Codex helped me find the right way to adapt rare switching to the private setting. The

Whose Good, Whose Place? The Moral Geography of Agentic AI for Social Good

arXiv:2605.22995v1 Announce Type: cross Abstract: Agentic AI systems are increasingly proposed for social-good domains, often invoking the United Nations Sustainable Development Goals (SDGs) as a voca

woah @Noahpinion is attacking @demishassabis here — calling what Demis said not “credible” — but misinterpreting what Hassabis is actually…

woah @Noahpinion is attacking @demishassabis here — calling what Demis said not “credible” — but misinterpreting what Hassabis is *actually* saying. What Demis actually said is both completely credibl