AICoder

AI Coding Live News

First-hand AI coding updates from 𝕏/Twitter frontier labs and researchers, distilled by AI with entity benchmarks.

Explore live on 𝕏 ↗

AI Coding Live News

H
Hugging Face Daily Papers@HuggingFace·
👀120
🔥98

GroundingPI Debuts: 4B Grounding Foundation Model for Physical Intelligence Outperforms GPT-6 Astra Across 34 Benchmarks

Robotic manipulation and physical intelligence systems suffer when relying on general-purpose vision backbones that struggle with dense clutter, tiny objects, and closed-loop control speeds. GroundingPI establishes a dedicated 4B grounding foundation model that unifies points and bounding boxes as quantized coordinates within a shared vocabulary. Outperforming 44 baselines across 34 benchmarks with an average of 73.68% (surpassing GPT-6 Astra at 71.54%), GroundingPI sets a new standard for embodied robotic visual backbones.

GroundingPI Debuts: 4B Grounding Foundation Model for Physical Intelligence Outperforms GPT-6 Astra Across 34 Benchmarks

Key Takeaways

  • Perceptual foundation for physical AI: 4B parameters, unifying visual points and boxes as quantized coordinates for real-time closed-loop control
  • Dominates 34 benchmarks: achieves 73.68% average score across 11 perceptual capabilities, surpassing the larger GPT-6 Astra (71.54%)
  • Empowers robotics and autonomous driving: outperforms existing backbones by up to 24.8% on RoboTwin 2.0 OOD and cuts required training demonstrations by half on RoboCasa-GR1
Read full story
Linked Models & Agents:
⚡ GroundingPI-4B⚡ GPT-6 Astra⚡ RoboCasa-GR1🤖 RoboTwin 2.0🤖 Embodied Agents🤖 VLA Models🌐 Hugging Face🌐 Physical Intelligence Research
#GroundingPI#Physical Intelligence#Embodied AI#Robotics#VLA#Computer Vision
H
Hugging Face Daily Papers@HuggingFace·
👀87
🔥99

Video2Skill: Streaming Embodied Skill Discovery from Video Feeds Exposes Bottlenecks in Open-World Skill Expansion

Embodied manipulation policies generalize across diverse scenes by recombining a core library of reusable physical skills. However, current agents rely almost exclusively on human-curated, hard-coded skill primitives. To achieve autonomous lifelong learning, agents must perform Streaming Embodied Skill Discovery (SESD)—observing uncurated continuous video streams and incrementally building a persistent skill catalog. Researchers from Northwestern University and CMU introduce Video2Skill (arXiv:2609.36691), benchmarking SESD across robotic manipulation and human kitchen tasks under three core capabilities: temporal event localization, physical transformation grouping, and the meta-decision to reuse existing skills versus instantiating novel primitives. Evaluating 19 open-source Vision-Language Models (VLMs) reveals severe structural limitations: most models cluster manipulation events at near-chance accuracy, and scaling parameters fails to resolve error rates. Furthermore, trained models suffer from skill library stagnation, consolidating familiar primitives while failing to expand into unobserved transformations. The authors propose Counterfactual Library-State Rebalancing (CLaRe) to mitigate clustering collapse (andyzworks.github.io/video2skill/).

Key Takeaways

  • Northwestern and CMU formulate Streaming Embodied Skill Discovery (SESD) and introduce Video2Skill benchmark across robot and human videos
  • Evaluates 19 open VLMs showing near-chance clustering accuracy; joint perception merges distinct skills while text pipelines duplicate them
  • Uncovers library stagnation: models fail to create novel skill primitives for unseen transformations, capping library size at <50% reference
Read full story
Linked Models & Agents:
⚡ Video2Skill⚡ CLaRe⚡ SESD🤖 Embodied-Skill-Agent🤖 Streaming-Discovery-Agent🤖 Lifelong-Learner🌐 Northwestern University🌐 Carnegie Mellon University🌐 Hugging Face
#Video2Skill#Embodied AI#Skill Discovery#Robotics#VLM#Lifelong Learning
H
Hugging Face Daily Papers@HuggingFace·
👀123
🔥99

ProgressCompass: Context-Aware Embodied Progress Reward Models Drive Long-Horizon Agent Planning

As embodied robotics agents tackle long-horizon tasks across household manipulation and industrial assembly, standard terminal outcome reward models (ORMs) fail to provide actionable learning signals along multi-step trajectories. While Process Reward Models (PRMs) evaluating step-by-step task progress percentages offer a viable solution, researchers from Northwestern University and CMU demonstrate in arXiv:2609.36684 that progress reward models collapse into random noise without appropriate context. The authors introduce ProgressCompass, formalizing the essential context dependencies required for embodied progress evaluation: task objectives, initial environmental baseline states, and continuous action history. Without grounding in preceding action sequences, PRMs routinely misjudge progress direction, confusing partial object states with completed goals. ProgressCompass provides an empirical testbed and modeling architectures to guide search and policy refinement in long-horizon robotic manipulation (andyzworks.github.io/progresscompass/).

Key Takeaways

  • Northwestern and CMU present ProgressCompass, formalizing essential context dependencies for embodied Process Reward Models
  • Demonstrates that progress estimation collapses without joint access to goal specs, initial baselines, and temporal action histories
  • Boosts long-horizon robotic task success by 38.6% and reduces interaction overhead by 41.2% via context-aware PRM trajectory guidance
Read full story
Linked Models & Agents:
⚡ ProgressCompass⚡ Process Reward Models⚡ Embodied PRM🤖 Embodied-Agent🤖 Progress-Evaluator🤖 Robotic-Planner🌐 Northwestern University🌐 Carnegie Mellon University🌐 Hugging Face
#ProgressCompass#Embodied AI#Process Reward Models#Robotics#Long-Horizon Planning#PRM
H
Hugging Face Daily Papers@HuggingFace·
👀95
🔥98

MotorMind Harness Scaffolds General Vision-Language Models for Zero-Shot Robotic Manipulation with 95% Real-World Success

Autonomous robot manipulation has long relied on specialized vision-language-action (VLA) models whose rigid training precludes zero-shot adaptation to novel environments, cutting them off from rapid foundation VLM advancements. Conversely, recent robotic agents rely heavily on fragile coding agents and auxiliary grounding models like SAM3, introducing severe latency and operational brittleness. Researchers from UIUC and CMU present MotorMind (arXiv:2609.38078), a direct robotic harness connecting general-purpose VLMs to deterministic robot control. By translating VLM mid-level actions into deterministic Cartesian primitives paired with asynchronous execution monitoring and background memory updates, MotorMind achieves 66.7% zero-shot success on LIBERO-PRO suites (versus 13.3% for prior baselines) and achieves 95% average success on a physical xArm6 robot under live human disturbances.

MotorMind Harness Scaffolds General Vision-Language Models for Zero-Shot Robotic Manipulation with 95% Real-World Success

Key Takeaways

  • Empowers general-purpose foundation VLMs to perform zero-shot robot manipulation without specialized VLA training or SAM3
  • Introduces mid-level action scaffolding paired with asynchronous monitoring for adaptive human-like disturbance recovery
  • Surges LIBERO-PRO benchmark success from 13.3% to 66.7% and achieves 95% success on a physical xArm6 robotic manipulator
Read full story
Linked Models & Agents:
⚡ MotorMind⚡ xArm6⚡ VLM🤖 Robotic-Manipulation🤖 Zero-Shot-Robot🤖 Mid-Level-Action🌐 UIUC🌐 Carnegie Mellon University🌐 Hugging Face
#MotorMind#Robotics#VLM#Zero-Shot Manipulation#Embodied AI
ADSponsored
G
George Yakushev, Denis Mazur et al.@HuggingFace·
👀46
🔥96

LLMs are General Asynchronous Agents: Non-Blocking Coroutine Runtimes Enable Concurrent Execution on Qwen 3.x Without Retraining

Mainstream LLM agents strictly follow sequential interaction loops: read, think, reply, call tools, and block until completion. Real-world applications—including full-duplex voice assistants, robotic manipulation, and continuous infrastructure monitoring—are inherently asynchronous, receiving incoming observations while the agent is actively deliberating or waiting on tool I/O. Researchers introduce a general asynchronous LLM framework that abstracts model inference into non-blocking coroutines with overlapping memory states. Without requiring task-specific fine-tuning or architectural alterations, foundation models like Qwen 3.x exhibit native asynchronous competency, effortlessly handling live video feeds, real-time gaming, and continuous systems monitoring.

LLMs are General Asynchronous Agents: Non-Blocking Coroutine Runtimes Enable Concurrent Execution on Qwen 3.x Without Retraining

Key Takeaways

  • Transition from Sequential Blocking to Asynchronous Coroutines: Establishes an inference runtime treating LLM reasoning as concurrent coroutines with overlapping memory states, breaking sequential execution limits.
  • Zero-Finetuning Asynchronous Competency: Empirical evaluations demonstrate that off-the-shelf Qwen 3.x models natively handle concurrent state synchronization and interrupts without bespoke fine-tuning.
  • Versatile Streaming & Real-Time Performance: Proves robust non-blocking adaptability across live video analysis, real-time game control, and concurrent infrastructure monitoring.
Read full story
Linked Models & Agents:
⚡ Qwen-3.5⚡ Qwen-2.5⚡ Llama-3.1🤖 Asynchronous Agent🤖 Streaming Agent🤖 Coroutine Agent🌐 Qwen🌐 Hugging Face🌐 arXiv
#Asynchronous#Coroutines#Multi-Task#Streaming#Agent Architecture#Concurrency
S
SWE-bench@SWEbench·
👀134
🔥92

EmbodiedSWE Benchmark Released: First Code-Agent Evaluation Suite for Long-Horizon Dexterous Robotics

The SWE-bench initiative and robotics researchers (@SWEbench) unveiled EmbodiedSWE (arXiv: 2609.27308). Expanding benchmark standards beyond pure software repositories, EmbodiedSWE assesses AI coding agents on 250 challenging robotic manipulation tasks requiring physical reasoning, sensor calibration, and closed-loop motor control code generation.

EmbodiedSWE Benchmark Released: First Code-Agent Evaluation Suite for Long-Horizon Dexterous Robotics

Key Takeaways

  • Features 250 enterprise-grade robotics software engineering challenges across ROS2, Isaac Gym, and MuJoCo.
  • Implements closed-loop physics simulation verification where candidate code must successfully execute in simulator.
  • Reveals that leading frontier models resolve under 22% of tasks, underscoring the gap between software and physical control.
  • Complete benchmark dataset and dockerized evaluation harnesses are openly accessible on GitHub and Hugging Face.
Read full story
Linked Models & Agents:
⚡ Claude 3.5 Sonnet⚡ DeepSeek-V3🤖 embodied-coder🌐 swe-bench
#SWE-bench#Embodied AI#Robotics#ROS2#Isaac Gym#Coding Agents
You've caught up with all stories