All entries

TTGBench: Benchmarking Topological Evolution and Semantic Drift in Text-attributed Temporal Graphs

arXiv:2609.08226v1 Announce Type: new Abstract: Temporal graph learning models the evolution of dynamic systems, where both structural interactions and semantic states change over time. However, exist

Content type

All Blog X Post Paper YouTube Reddit GitHub

Safety

Two big updates 1. I published an @FT op-ed on the OpenAI/HF incident & follow-ups 2. We’re starting an Open Alignment team at @huggingface …

DGX agent

Two big updates 1. I published an @FT op-ed on the OpenAI/HF incident & follow-ups 2. We’re starting an Open Alignment team at @huggingface to work on safety & alignment for open models, incl cybersec

safetyclem-delangue--x

10 Sept 2026

Two-Token Features and Small-Large Ensembles for VLM Hallucination Detection

arXiv:2609.10244v1 Announce Type: new Abstract: We present our system for the SHROOM-Visions 2026 shared task on character-level VLM hallucination detection. A small (4B-parameter) VLM is fine-tuned a

Typed Federated Artifacts for the Agentic Web:Sharing Tool-Routing Knowledge Across Frozen,Heterogeneous LLM Agents

arXiv:2609.06815v1 Announce Type: cross Abstract: An open, networked web will allow agents to run frozen models from multiple vendors, keep their history private, and teach each other which tool to ca

Industry

Typewise orchestrates customer-service AI agents

Swiss artificial intelligence startup Typewise AG today launched Nova, a system designed to build, run and continuously improve customer-service AI agents without requiring dedicated operations specia

industrysiliconangle

Tutorials

uFlowCSP: Crystal Structure Prediction using Mean flow generative models

arXiv:2609.09799v1 Announce Type: cross Abstract: Crystal structure prediction (CSP) is fundamental to computational materials discovery. Generative models including CDVAE, DiffCSP, FlowMM, and Crysta

tutorialsarxiv-cs-lg

Unbiased and Biased Variance-Reduced Forward-Reflected-Backward Splitting Methods for Stochastic Composite Inclusions

arXiv:2603.15576v2 Announce Type: replace Abstract: This paper develops new variance-reduction techniques for the forward-reflected-backward splitting (FRBS) method to solve a class of possibly nonmon

safetyarxiv-cs-lg

Uncertainty-Aware Sea-Ice Type Mapping with Multiple Ice Charts

arXiv:2609.09451v1 Announce Type: new Abstract: Sea-ice stage of development (SoD) describes the age and associated thickness of sea ice and provides important information for navigation, and operatio

researcharxiv-cs-lg

Understanding the Impact of Model Pruning on Long-Tail Forgetting and Explanation Reliability in Medical Imaging

arXiv:2609.07803v1 Announce Type: new Abstract: Model pruning is widely used to compress deep neural networks, reducing memory and computational requirements with minimal impact on aggregate performan

UniRRM: Unified Reasoning Reward Models Across Languages and Evaluation Paradigms

arXiv:2609.05910v1 Announce Type: cross Abstract: Reinforcement learning (RL) excels on tasks with verifiable rewards, but in open-ended tasks, the reliability of reward models remains a key challenge

UnitBoost: Managing Compound LLM Systems with a Merge Operator, Not a Model

arXiv:2609.09815v1 Announce Type: cross Abstract: Compound LLM systems often solve a coordination problem by adding a higher-level LLM. The resulting meta-agent reads workers' outputs, writes the fina

agentsarxiv-cs-cl

Unsound Search with Policy and Value Networks in Legends of Code and Magic

arXiv:2609.06816v1 Announce Type: new Abstract: Decision-time search in perfect and imperfect information games with enumerable belief states are effective methods for game AI. Collectible card games

safetyarxiv-cs-ai

Unthrottling the Tanh Jacobian in SAC: A Negative Result on Bang-Bang Control and MetaDrive

model-releasesarxiv-cs-lg

arXiv:2609.09478v1 Announce Type: new Abstract: Soft Actor-Critic (SAC) represents a continuous policy as an unbounded Gaussian that is squashed by tanh. The Jacobian of that map is partial a/partial

Unveiling Hidden Threats: Using Fractal Triggers to Boost Stealthiness of Distributed Backdoor Attacks in Federated Learning

arXiv:2511.09252v2 Announce Type: replace-cross Abstract: Traditional distributed backdoor attacks (DBA) in federated learning improve stealthiness by decomposing global triggers into sub-triggers, wh

UOT-Gap: A Variational Principle for the Modality Gap in Vision-Language Models via Unbalanced Optimal Transport

arXiv:2609.10224v1 Announce Type: new Abstract: Vision-language models such as CLIP embed images and text in a shared space, where modality-specific distributions often remain separated. Existing acco

safetyarxiv-cs-cv

Vague2Detect: Handling Ambiguous Prompts in Knowledge-Based Open-World Detection

arXiv:2609.09949v1 Announce Type: cross Abstract: Real-world detectors must often interpret functional or ambiguous prompts, yet conventional models such as YOLO remain restricted to fixed class lists

VANTAGE-Bench: Evaluating the Infrastructure AI Gap in Vision-Language Models

model-releasesarxiv-cs-cv

arXiv:2609.09396v1 Announce Type: new Abstract: As Vision-Language Models (VLMs) advance toward physical deployment, the focus has remained on action-oriented Embodied AI evaluated on subject-centric

Versioned Transitive Dependency-Closure Binding and Operation-Time Effect Governance for Agent Skills: ClosureBound

arXiv:2609.05920v1 Announce Type: cross Abstract: Agent Skills combine instructions with files, packages, tools, models, and services, so operational identity can exceed a signed directory. Recursive

agentsarxiv-cs-ai

VFNet: Multi-View Spatio-Temporal Model for Void Fraction Estimation in Gas-Liquid Two-Phase Flow

model-releasesarxiv-cs-cv

arXiv:2609.09711v1 Announce Type: new Abstract: Void fraction, which quantifies the proportion of the fluid flow volume occupied by the gas phase, is a key parameter in the characterization of gas-liq

ViBe: Visual Behavior Adaptation for Perceptive Humanoid Whole-Body Control

model-releasesarxiv-cs-ro

arXiv:2609.09918v1 Announce Type: new Abstract: Motion tracking provides a scalable recipe for humanoid whole-body control. By design, the resulting trackers lack exteroceptive feedback hence reacting

Tutorials

Video and image search in Amazon Bedrock Knowledge Base using Marengo 3.0

TwelveLabs Marengo Embed 3.0 is now generally available as an embedding model in Amazon Bedrock Knowledge Bases, bringing fully managed natural language search to video, image, and audio content. This

tutorialsaws-ml-blog

Video-MOPD: Multi-Teacher On-Policy Distillation for Video Understanding

arXiv:2609.09300v1 Announce Type: new Abstract: Video understanding demands a convergence of complementary capabilities across perception, temporal understanding, and complex reasoning, which are diff

safetyarxiv-cs-cv

VideoTIR: Accurate Understanding for Long Videos with Efficient Tool-Integrated Reasoning

arXiv:2603.25021v3 Announce Type: replace Abstract: Existing Multimodal Large Language Models (MLLMs) often suffer from hallucinations in long video understanding (LVU), primarily due to the imbalance

safetyarxiv-cs-cv

View-Structured Conformal Prediction for 3D Gaussian Splatting

arXiv:2609.10307v1 Announce Type: cross Abstract: 3D Gaussian Splatting (3DGS) renders novel views in real time, but an uncertainty heatmap does not certify that a rendered view meets a certain predic

researcharxiv-cs-cv

Vision-language models know more about agriculture than they show and rubric-grounded verifications close the gap

model-releasesarxiv-cs-cv

arXiv:2609.09417v1 Announce Type: new Abstract: Vision-language models (VLMs) show promise for agricultural classification, but zero-shot performance on disease, pest, damage, quality, and species ide

Local Ai

VLLM-MLX vs Ollama MLX... Ran concurrency benching. Qwen3.5-4B.

Anyone else play with vllm-mlx? There was an interesting paper that came out that compared this version of vllm that was originally designed for multi modal inference performance gains, but ended up s

local-air-ollama

VLX-VR: An Agentic-Aware Video Reasoning Model

arXiv:2609.09985v1 Announce Type: new Abstract: Real-world video understanding requires integrating visual, audio, textual, and temporal evidence distributed across a video. Yet many pipelines use a f

agentsarxiv-cs-cl

VST: Verifiable Structured Transport for Auditable Agent-to-Agent Alpha Discovery

arXiv:2609.07065v1 Announce Type: new Abstract: Agent-to-agent (A2A) alpha discovery is slowed by repeated feedback cycles between mining and evaluation agents, whose hand-offs, in contemporary LLM mu

WAPP: Safe Learning of Positive Security WAF Policies from Live Traffic

arXiv:2609.06840v1 Announce Type: cross Abstract: Web Application Firewalls (WAFs) mainly rely on signatures to detect known attacks, which can leave gaps against modified or previously unseen payload

safetyarxiv-cs-ai

We Built a Mirror and Mistook It for a Mind: Causal Liability and the Fallacy of AI Consciousness

arXiv:2609.06715v1 Announce Type: new Abstract: The contemporary debate over machine consciousness begins from a concealed assumption: that the object called 'AI' already constitutes the kind of entit

We're Cooked! - Probing LLM Political Alignment Via Conflict-Framed Recipe Translation

arXiv:2609.07568v1 Announce Type: cross Abstract: Large language models (LLMs) are increasingly deployed for translation tasks, yet their implicit political positioning in such contexts remains unders

We're publishing our most detailed threat intelligence report to date. It covers how people tried to misuse Claude—for cyberattacks, influen…

model-releaseskarpathy--x

We're publishing our most detailed threat intelligence report to date. It covers how people tried to misuse Claude—for cyberattacks, influence operations, surveillance, biology, and building weapons—a

We've trained new models for checkbox parsing ☑️ This lets you convert checkboxes of any size and shape, filled or unfilled, into structured…

We've trained new models for checkbox parsing ☑️ This lets you convert checkboxes of any size and shape, filled or unfilled, into structured JSON that's suitable for programmatic or agentic use. Use i

agentsjerry-liu--x

What Does a System Modify When It Modifies Itself?

arXiv:2603.27611v2 Announce Type: replace Abstract: When a cognitive system modifies its own functioning, what exactly does it modify: a low-level rule, a control rule, or the criterion that evaluates

What Does an LLM-Agent Leaderboard Rank Actually Compare?

arXiv:2609.07785v1 Announce Type: new Abstract: An LLM-agent leaderboard invites a familiar inference: an agent ranked above another is the better agent. Public evaluation logs may not support that co

What Does MMLU Actually Measure? A Psychometric Audit of Difficulty Structure in Aggregate Benchmark Scores

arXiv:2609.09372v1 Announce Type: cross Abstract: Although MMLU is widely adopted as a benchmark for calibrating general AI capabilities, we psychometrically demonstrate that its aggregate score prima

What Eviction Destroys: A Restore-Counterfactual Audit of Forgetting in Agent Memory

arXiv:2609.08279v1 Announce Type: cross Abstract: Agent memory systems must discard stored information when their history exceeds a fixed token budget. Existing budget-accuracy frontiers quantify the

What Fixed-Rollout pass@k Evaluations Can Identify

arXiv:2609.09245v1 Announce Type: cross Abstract: Repeated-sampling evaluations increasingly extrapolate pass@k far beyond the number n of samples collected per problem. We show that, in the pooled/ra

researcharxiv-cs-lg

Applications

What if LLMs Ate Their Words: Causal History Effects in Multi-Turn Interaction

arXiv:2609.05882v1 Announce Type: cross Abstract: Multi-turn interaction creates a feedback process in which an LLM's previous responses become context for later behavior. Prior work shows substantial

applicationsarxiv-cs-ai

What LLM Trading Agents Actually Do in Production: A Six-Month, Population-Scale Record from Two Fleets

arXiv:2609.05663v1 Announce Type: new Abstract: We present a continuous, population-scale measurement record of autonomous language-model trading agents operating in production across two systems with

What Makes Adversarial Examples Transfer Across Deepfake Detectors?

arXiv:2609.10002v1 Announce Type: new Abstract: Deepfake detectors remain vulnerable to transfer-based black-box attacks, in which adversarial examples are generated on a source surrogate model and tr

researcharxiv-cs-cv

What Symmetry Buys a Learned Motion Planner

model-releasesarxiv-cs-ro

arXiv:2609.10033v1 Announce Type: new Abstract: Learning-based motion planners pay at training what classical planners pay per query. Trained in world coordinates, they relearn the same motion at ever

What to run at 128GB VRAM?

model-releasesr-localllama

Long time lurker, but I'm finally upgrading to 128GB VRAM, and I'm trying to figure out what to run. I had been leaning towards Qwen3.8 Flash-Next at ~Q4, and I generally prefer to not run anything be

When Agent Governance Helps

arXiv:2609.05531v1 Announce Type: cross Abstract: No specification says how a governed autotelic AI agent organization, where agents pursue self-generated goals inside guardrails, should be designed a

When and Why LLM Causal Priors Help: Closed-Loop Prior Selection for Amortized Causal Inference

arXiv:2609.06941v1 Announce Type: new Abstract: Causal effect estimation asks how an outcome would change under an intervention, and medicine, economics, and public policy all treat it as a foundation

When Auditors Fabricate: Batch-Size Degradation and Confident Hallucination in LLM Detection of Planted Document Contamination

arXiv:2609.09696v1 Announce Type: new Abstract: Large language models are increasingly proposed as automated auditors of document quality, yet their reliability as detectors of planted errors is poorl

When Can LLM Digital Twins Reduce Human Measurement? From Behavioral Fidelity to Statistical Substitutability

arXiv:2609.07987v1 Announce Type: new Abstract: LLM-based digital twins promise to reduce repeated human data collection by generating person- specific responses, yet existing evaluations provide litt

When Can One Obtain Certificates of Optimality Using Positivstellensaetze?

arXiv:2609.08736v1 Announce Type: new Abstract: We study certificates of positivity and optimality for learning problems whose objectives and constraints need not be polynomial. We isolate an axiomati