Skip to content
AI News HubLIVE

This edition’s highlights

Agents

CodeRabbit scales agentic software development and ships 20x more

  • CodeRabbit built its AI-powered code review and governance platform using OpenAI models from GPT-3.5 Turbo to GPT-6 series.
  • With Codex, CodeRabbit has shipped roughly 20 times more software over the last six months without adding engineers.
OpenAI NewsIn-site articleCodeRabbit scales agentic software development and ships 20x more

Hide your slop PR descriptions behind

  • GitHub Markdown supports , letting PR descriptions collapse long text behind an expandable summary line
  • Agent-written PR descriptions are accurate and comprehensive but read like AI slop, prompting a reflexive "I'm not reading all of that"
Simon Willison's WeblogIn-site articleHide your slop PR descriptions behind
Research

Kuken Industries searches 70 years of drawings in minutes

  • Time to find a drawing fell from nearly an hour to 4.5 minutes.
  • Kuken projects a 90% reduction in hours spent searching for drawings each year.
OpenAI NewsIn-site articleKuken Industries searches 70 years of drawings in minutes

From Geometry to Function: A Multi-Task LiDAR Benchmark Dataset for Semantic and Functional Parsing of Urban Facades

  • HFX3D contains about 492 million LiDAR points from 26 Halifax buildings across four seasons, day and night, and varied weather.
  • Each facade is annotated at semantic, instance, and per-instance functional attribute tiers, claimed as the first facade dataset pairing geometry and semantics with functional labels.
arXiv Computer VisionIn-site articleFrom Geometry to Function: A Multi-Task LiDAR Benchmark Dataset for Semantic and Functional Parsing of Urban Facades

parse_number() as a SQLite extension: C vs Rust

  • Rust extensions were 298–321 KB versus 14 KB for the C extension
  • Plain C was slow on decimals due to strtod, but an optimized fast path closed the gap
Simon Willison's WeblogIn-site articleparse_number() as a SQLite extension: C vs Rust
Chips

Accelerate GPU training on Unity Catalog tables with UCDataset

  • UCDataset is a native PyTorch data loader for Unity Catalog that streams table data directly to GPUs, eliminating export and copy steps.
  • It achieves a 10x speedup in training time compared to loading from Parquet files on 8 NVIDIA H100 GPUs for a recommendation model.
Databricks BlogIn-site articleAccelerate GPU training on Unity Catalog tables with UCDataset
Models

Towards Label-Efficient Geospatial Foundation Model Adaptation via Positive-Unlabeled Learning for Building Damage Mapping

  • PUMA requires annotations only for a small number of damaged buildings.
  • It integrates non-contrastive weakly supervised representation refinement and latent-state label disambiguation without class prior estimation.
arXiv Computer VisionIn-site articleTowards Label-Efficient Geospatial Foundation Model Adaptation via Positive-Unlabeled Learning for Building Damage Mapping

ActPrune: Action-Grounded Visual Token Pruning for Efficient Vision-Language-Action Models

  • ActPrune moves the pruning decision ahead of the vision encoder by grounding it in the action chunk predicted at the previous control step.
  • It applies asymmetric pruning to the third-person and wrist views: retaining end-effector trajectory patches and semantic anchors in the fixed view, while pruning unchanged patches after camera-motion compensation in the wrist view.
arXiv RoboticsIn-site articleActPrune: Action-Grounded Visual Token Pruning for Efficient Vision-Language-Action Models
Robotics

What Limits mmWave Human-State Sensing for Robot Perception?

  • The framework models the human and environment as a time-varying scattering field and states an error floor for joint localization.
  • The floor depends on bandwidth, aperture, SNR, fused frame count, phase coherence, and visibility of retained returns.
arXiv RoboticsIn-site articleWhat Limits mmWave Human-State Sensing for Robot Perception?

You have reached the end of this edition.

Your reading list
Other updates (35)
Robotics

Practice Makes Precise: At CoRL, NVIDIA and Partners Reveal How Robots Rehearse Delicate Work in Simulation

NVIDIA unveiled GR00T 2 world action model, Isaac Lab 3.0, Newton physics integration, DexBench and VABAR benchmarks, and agentic robotics workflows at CoRL, advancing robots from locomotion to precise manipulation with partners like Ambi Robotics, Bosch, and Siemens.

NVIDIA BlogIn-site articlePractice Makes Precise: At CoRL, NVIDIA and Partners Reveal How Robots Rehearse Delicate Work in Simulation

STRIDE: Learning Humanoid Locomotion in Minutes with Acceleration-Based Rewards

A new arXiv paper, STRIDE, proposes a reinforcement learning reward that relies only on the onboard IMU by tracking a commanded acceleration profile instead of base velocity, enabling Unitree G1 and Booster T1 humanoids to learn to walk within about 14 minutes of samples while sharply reducing training falls.

arXiv RoboticsIn-site articleSTRIDE: Learning Humanoid Locomotion in Minutes with Acceleration-Based Rewards

ForPC: Force-Aware Whole-Body Control for Precise Humanoid Cart Pushing

Researchers propose ForPC, a force-aware whole-body control framework that regulates cart velocity via two-stage reinforcement learning. Without external force sensors or human demonstrations, it significantly outperforms a locomotion-only baseline in simulation and transfers zero-shot to a physical Unitree G1, pushing a 30 kg cart at 0.5 m/s.

arXiv RoboticsIn-site articleForPC: Force-Aware Whole-Body Control for Precise Humanoid Cart Pushing

BEVCast: Path-Adaptive BEV for Efficient End-to-End Autonomous Driving

BEVCast introduces a two-stage bird's-eye-view representation whose resolution adapts to the predicted driving path: a coarse 4.096 m grid encodes the whole scene, and only the highest-heat cells are refined to 1.024 m under a fixed budget, cutting BEV encoder time to 66% while improving accuracy on nuScenes and NAVSIM.

arXiv Computer VisionIn-site articleBEVCast: Path-Adaptive BEV for Efficient End-to-End Autonomous Driving
Agents

Import AI 480: Self-sustaining space economies; AI for science progress, and Meta builds perfect hands for a superintelligence

This issue covers an AI startup proving a quantum conjecture, Meta's Tactus gloves for dexterous manipulation data, a NASA paper on affordable lunar industrialization, and Core Automation's distributed training efforts.

Import AIIn-site articleImport AI 480: Self-sustaining space economies; AI for science progress, and Meta builds perfect hands for a superintelligence

It Takes Two to Tangram: Multi-Agent Object Centric Reinforcement Learning for Assembly

This paper presents an object-centric reinforcement learning framework that treats each part as an agent and learns a shared policy for geometric assembly. Using a graph neural network to encode parts, target poses, and spatial relationships, combined with per-agent PPO and a centralized graph critic, the method achieves concurrent assembly in simulation on planar Tangram tasks and transfers to sequential execution on a real robot.

arXiv RoboticsIn-site articleIt Takes Two to Tangram: Multi-Agent Object Centric Reinforcement Learning for Assembly

An Anatomy-guided 3D Segmentation Pipeline for Prostate Surgical Phantoms from CT Scans

This work presents an automated 3D segmentation pipeline that uses classical medical image-processing techniques to encode known phantom anatomy as explicit geometric and intensity priors, eliminating the need for training data. Across 57 scans, it achieved 0.83 median Dice on 11 contrast-suppressed scans, detected all 32 annotated focal lesions without false positives, and completed scans in 2-8 minutes on CPU versus 1-2 hours manually.

arXiv Computer VisionIn-site articleAn Anatomy-guided 3D Segmentation Pipeline for Prostate Surgical Phantoms from CT Scans

From Procedures to Permissions: Learning Precise Subagent Authorization from Agent Skills

This paper proposes using procedural knowledge encoded in Agent Skills as supervision to precisely determine tool and workspace permissions for subagents. Experiments show significant improvements in permission precision and task success on ClawArena-Team.

arXiv Computational LinguisticsIn-site articleFrom Procedures to Permissions: Learning Precise Subagent Authorization from Agent Skills

FinEvolveKG: Controlled and Attributable Prompt Evolution for Financial Knowledge Graph Construction

FinEvolveKG introduces a knowledge graph construction harness that separates reusable general rules from persona policies and mechanically confines prompt optimization to designated blocks, improving financial KG quality while keeping changes reviewable and attributable.

arXiv Computational LinguisticsIn-site articleFinEvolveKG: Controlled and Attributable Prompt Evolution for Financial Knowledge Graph Construction
Research

Control Room: A central system to track and stop model workloads across our research infrastructure

OpenAI introduces Control Room, a centralized system to track model workloads across research infrastructure and rapidly halt model output within 35 seconds when safety or security concerns arise. The system is designed to fail closed and is resilient to buggy workload code.

OpenAI NewsIn-site articleControl Room: A central system to track and stop model workloads across our research infrastructure

CS-Pi: Constraint-Satisfying Flow Policy Sampling

CS-Pi introduces a sampling method that enforces novel constraints at inference time for flow-matching policies by combining projection efficiency with rejection-sampling plausibility. It achieves 81% safe success on a 7-DoF shelf task and 55% on a zero-shot Franka deployment, outperforming projection baselines.

arXiv RoboticsIn-site articleCS-Pi: Constraint-Satisfying Flow Policy Sampling

Q-OIL: Optimistic Exploration in Learning from Human Interventions

The paper shows that common human-intervention modeling assumptions do not match real behavior, introduces a progress-based intervention model, and proposes Q-OIL, which adds optimistic bonuses to intervention transitions with decoupled critics, achieving 10-50% higher success in simulation and on real robots.

arXiv RoboticsIn-site articleQ-OIL: Optimistic Exploration in Learning from Human Interventions

VoluCompose: Constructing Compact Volumetric Visual Representations for Medical Image Understanding

VoluCompose is a training-free, model-agnostic framework that decouples what visual content to retain from how many tokens to allocate, using hyperbolic token composition and attention-entropy-based capacity control to compress medical volumetric images. Across three datasets and four backbones, it retains 22.78% of visual tokens on average while reaching 121.68% mean relative quality versus uncompressed references.

arXiv Computer VisionIn-site articleVoluCompose: Constructing Compact Volumetric Visual Representations for Medical Image Understanding

HaNDF: Object-Conditioned Geometric Neural Hand Distance Fields

The paper introduces HaNDF, an object-conditioned neural distance field used as a data-driven prior for plausible hand-object configurations. Leveraging Riemannian geometry, it represents plausible hand poses as the zero-level set of an object-induced conditional field and can also optimize for the interacting object, supporting pose refinement, reconstruction under occlusion, and physically plausible manipulation synthesis. It was accepted at NeurIPS 2026.

arXiv Computer VisionIn-site articleHaNDF: Object-Conditioned Geometric Neural Hand Distance Fields

Semantic Conditional Similarity (SCS): A Formal Mathematical Framework

The paper introduces Semantic Conditional Similarity (SCS), a conditional form of STS that computes similarity between two documents after eliminating semantic elements shared with a third, extending pairwise STS to ordered triples with directionality and role-dependent asymmetry.

arXiv Computational LinguisticsIn-site articleSemantic Conditional Similarity (SCS): A Formal Mathematical Framework

Contextuality and Contradictions: Exploring Real-World Identifications of AI-Generated Writing

An arXiv study uses qualitative coding and computational analysis of lexical change to examine how English Wikipedia editors and students identify AI-generated text. It finds that such identifications draw on diverse and often contradictory evidence, are context-dependent, and are frequently motivated by 'pragmatic infelicity'—a perceived mismatch between textual elements and contextual expectations—rather than any specific feature. AI identification thus serves as a discursive resource for maintaining sociolinguistic norms, and detection tools that treat AI-likeness as context-free may inherit and enforce the norms of their training contexts at scale.

arXiv Computational LinguisticsIn-site articleContextuality and Contradictions: Exploring Real-World Identifications of AI-Generated Writing
Policy

Introducing Control Room: Human control over AI workloads

OpenAI introduces Control Room, a centralized system to track model workloads across research infrastructure and stop them quickly when safety or security concerns arise.

OpenAI NewsIn-site articleIntroducing Control Room: Human control over AI workloads

National and State-Level Approaches to Automated Vehicle Certification: A Comparative Framework for State Operational Governance

The paper analyzes how U.S. states regulate SAE Level 4-5 automated driving systems, proposes a three-model certification typology, compares practices in the EU, China, Japan, and Australia, and finds North Carolina currently follows a baseline compliance model.

arXiv RoboticsIn-site articleNational and State-Level Approaches to Automated Vehicle Certification: A Comparative Framework for State Operational Governance
Startups

NVIDIA Announces New Planned Investments in Korean Startups at the MSIT-NVIDIA AI Startup Partnership Forum

Korea's Ministry of Science and ICT and NVIDIA hosted the AI Startup Partnership Forum at COEX, where NVIDIA announced an investment in Point2 and planned investments in Motif Technologies, RLWRLD, and Upstage, covering the full AI stack from infrastructure to models to robotics applications.

NVIDIA BlogIn-site articleNVIDIA Announces New Planned Investments in Korean Startups at the MSIT-NVIDIA AI Startup Partnership Forum
Chips

Korea Advances Physical AI Across Industries With NVIDIA

NVIDIA is helping advance Korea's physical AI ecosystem by tapping into the country's manufacturing and engineering strengths. Industry leaders like Samsung, Kakao Mobility, and LG Electronics are applying accelerated computing, AI, and simulation to semiconductor manufacturing, robotics, and autonomous vehicles, bringing physical AI from development into real-world production.

NVIDIA BlogIn-site articleKorea Advances Physical AI Across Industries With NVIDIA

InstaGen: Group-Shared Importance Sampling for Fast Video Generation

InstaGen recasts block selection in permuted block-sparse attention (PBSA) as importance sampling: top-scoring keys are computed exactly, the rest are sampled in proportion to estimated attention and reweighted by inverse sampling probability, with one sample shared across each query cluster. Parameterized by SVG2, it improves PSNR by 3.4 dB and 2 dB over SVG2 on Wan2.1-T2V-1.3B and 14B at matched attention density, and by 1.6 dB and 2.2 dB at matched wall-clock time; on the 14B model it also beats SVG-EAR by 1.1 dB at similar density while running faster.

arXiv Computer VisionIn-site articleInstaGen: Group-Shared Importance Sampling for Fast Video Generation

NVIDIA Dynamo Recipes Bring Korea’s Leading Open Models to Production

NVIDIA releases Dynamo-optimized deployment recipes for Korea’s top open models. Optimized for B200 GPUs, the recipes deliver up to 2x performance out of the box, helping developers deploy models faster.

NVIDIA BlogIn-site articleNVIDIA Dynamo Recipes Bring Korea’s Leading Open Models to Production
Models

Korea’s AI Ecosystem Advances and Secures Agentic AI With NVIDIA

Hyundai Motor Group is exploring NVIDIA NemoClaw blueprints and the NVIDIA OpenShell runtime, while AIM Intelligence integrates its Starfort platform with OpenShell to support secure AI agents. LG AI Research is opening API access for K-EXAONE-2.5, developed using NVIDIA Nemotron datasets.

NVIDIA BlogIn-site articleKorea’s AI Ecosystem Advances and Secures Agentic AI With NVIDIA

VICTR: Value-Informed In-Context Learning for Vision-Language-Action Models

VICTR introduces progress-informed context retrieval for vision-language-action models via an In-Context Value Function Estimator (IC-VFE). By estimating task-relative progress from a reference demonstration, a language instruction, and the current observation, VICTR selects demonstration chunks at matching execution stages for a π0.5 policy post-trained for in-context action generation. On unseen out-of-domain real-world manipulation tasks, it raises success from 6.2% to 23.8% and task progress from 28.9% to 55.5%.

arXiv RoboticsIn-site articleVICTR: Value-Informed In-Context Learning for Vision-Language-Action Models

Meituan-Robotics-0: A Vision-Language-Action Foundation Model for Desktop Manipulation

Meituan Robotics introduces MR0, a vision-language-action foundation model and embodied-agent system for stationary-base desktop manipulation. It combines 147 million vision-language samples with 15,060 hours of robot trajectories across 12 embodiments, uses a three-stage training recipe, and proposes Rollback for CFG-DAgger. MR0 achieves 93.6% success on RoboTwin 2.0 and ranks fifth in the RoboDojo VLA track.

arXiv RoboticsIn-site articleMeituan-Robotics-0: A Vision-Language-Action Foundation Model for Desktop Manipulation

Turkish Multimodal Video Fact-Checking: A New Dataset and a Comparison of Zero-Shot Agentic and Fine-Tuned Systems

This paper introduces TeViD, the first multimodal video fact-checking dataset for Turkish, and compares two verification pipelines built on the same open-source base model: a zero-shot agentic pipeline (3MFact-TR) and a LoRA fine-tuned closed-book model, reporting accuracy, macro F1, and ablation studies.

arXiv Computer VisionIn-site articleTurkish Multimodal Video Fact-Checking: A New Dataset and a Comparison of Zero-Shot Agentic and Fine-Tuned Systems

Subliminal Learning in Vision-Language Models

A study accepted to WACV 2027 finds that vision-language models can transfer hidden preferences—such as a bias toward a specific animal or facial attribute—from a teacher model to a student model through an unrelated visual task involving geometric shapes and generated numbers. The transfer reaches up to about 10 percentage points in the animal preference task and also shifts how students explain their choices.

arXiv Computer VisionIn-site articleSubliminal Learning in Vision-Language Models

Richness Is Not Evenness: Ordered Renyi Crossings under Self-Training

This paper studies recursive self-training of language models and finds that as training temperature rises, different Renyi entropies (min-entropy, collision entropy, Shannon entropy, and observed richness) cross from loss to gain at distinct temperatures in a consistent order, which persists across various settings.

arXiv Computational LinguisticsIn-site articleRichness Is Not Evenness: Ordered Renyi Crossings under Self-Training

Guardrail Erosion in Domain Fine-Tuned Large Language Models: A Paired Audit of Knowledge-Centric Abstention in Biomedical LLMs

A paired audit of 25 biomedical fine-tuned models against 11 base checkpoints finds that appropriate knowledge-centric abstention on MedAbstain drops from a median of 8.3% in base models to 4.2% after fine-tuning, decreasing in 21 of 25 pairs with heterogeneous magnitude and direction.

arXiv Computational LinguisticsIn-site articleGuardrail Erosion in Domain Fine-Tuned Large Language Models: A Paired Audit of Knowledge-Centric Abstention in Biomedical LLMs

Structuring Survey Histories for LLM Digital Twins: A Controlled Study of Background, Decision Procedure, and Evaluation Profiles

A controlled study finds that organizing a respondent's survey history into a fixed Background–Decision procedure–Evaluation (BDE) profile improves LLM digital-twin prediction accuracy over raw survey transcripts, and that the gain stems mainly from profile composition rather than compression or deeper inference.

arXiv Computational LinguisticsIn-site articleStructuring Survey Histories for LLM Digital Twins: A Controlled Study of Background, Decision Procedure, and Evaluation Profiles

RL-ATTACK: A Generalized Black-Box Adversarial Attack on Language Models via Reinforcement Learning

Researchers propose RL-ATTACK, a trainable reinforcement learning framework for black-box adversarial attacks that learns from the target model to generate tailored adversarial transformations. It drives post-attack accuracy to 2% within 20 queries on natural language inference and outperforms an untrained attacker and a strong optimization baseline in generative Q&A. The paper is accepted at IEEE EdgeCom 2026.

arXiv Computational LinguisticsIn-site articleRL-ATTACK: A Generalized Black-Box Adversarial Attack on Language Models via Reinforcement Learning

Compression of Emotion in Language Model Representations

A study of seven LLMs on the GoEmotions dataset finds that hidden representations overcompress fine-grained emotion categories into coarser sentiment and emotion-family groupings, yet linear probes retain substantial fine-grained predictive information.

arXiv Computational LinguisticsIn-site articleCompression of Emotion in Language Model Representations

Forcing a Letter Suppresses Refusal: Mechanistic Evidence from WMDP

A study accepted to BlackboxNLP 2026 provides the first mechanistic explanation for why multiple-choice question format reduces refusal rates in LLMs: models commit quickly to answer letters, activations lie farther from the refusal direction, and answer selection competes with refusal.

arXiv Computational LinguisticsIn-site articleForcing a Letter Suppresses Refusal: Mechanistic Evidence from WMDP

The Web Is More Than Enough, by Paul Ford

Simon Willison, inspired by a Hacker News comment on Paul Ford's essay, used Claude Opus 5.5 to port a Three.js Nudibranch-themed Faberge Egg to vanilla JavaScript and WebGL without Three.js. The original is about 77KB and the port is about 98KB; the two are not identical. Willison finds the Three.js code slightly easier to read but says he rarely looks at the code for such 3D projects.

Simon Willison's WeblogIn-site articleThe Web Is More Than Enough, by Paul Ford