Grid Robotics paper pitches 'agentic robotics' powered by GPT-6 Astra
A Grid Robotics research team laid out a new paradigm called agentic robotics, pairing foundation models like GPT-6 Astra with an "embodied harness" so machines can reason, act and verify in the physical world. The arXiv paper, out November 13, 2026, drew 107 upvotes on HuggingFace Daily Papers within days.
- Original source: blockchain.news.
- The source published it on Nov 17, 2026 at 3:53 AM.
- No second outlet has reported it yet. Treat it as a single-source claim.
The paper landed on arXiv on November 13, 2026, with Grid Robotics researchers Qingfeng Lan, Duo Li and Siyao Li leading the work. By November 16 it had collected 107 community upvotes on HuggingFace Daily Papers. The pitch: stop hand-building task-specific policies and let a general foundation model drive the robot instead.
That core is what the authors call the Embodied System 2 — a foundation model fused with an embodied harness to handle the deciding. It runs a closed loop: watch the environment, read user intent, reason about the task, call up the right skills and tools, check the result, then switch strategy when something goes sideways. Traditional robotics stacks, by contrast, lean on narrow policies built for one job.
Reality bites hard, and the paper says so. Turning general reasoning into actions you can actually verify runs into limited reaction times, mechanical limits and environments that refuse to behave. Those constraints are why the authors push for a shared evaluation framework and a common research roadmap.
Their proposed framework scores two things at once: what the foundation model can do, and how the whole closed loop performs. It's aimed squarely at the field's missing piece — no unified definition and no reliable benchmarks for agentic robotics exist yet.
The roadmap sketches a climb from executing single primitives up to holding down long-term roles with wider responsibilities, safely and reliably even when conditions get dynamic and uncertain. The team's goal is to carry principles that work for digital agents over into physical ones, closing the gap between theory and real machines.
GPT-6 Astra is the headline engine here, but the authors tie success to the community converging on standard definitions, evaluation metrics and joint research. Without that, sustained physical autonomy stays a paper promise.
TopicsGrid Robotics · GPT-6 Astra · Embodied System 2 · arXiv · HuggingFace Daily Papers · Qingfeng Lan · Duo Li · Siyao Li
Source: blockchain.news
Written by AI from verified reporting.
More stories
- Tech & AI x.com 0 viewsNvidia director warns board of losing "irreplaceable talent" if Huang isn't reinstated as CEO
- Tech & AI blockchain.news 0 viewsNVIDIA's Vera Rubin aces MLPerf debut with 10x agentic efficiency gain
- Tech & AI blockchain.news 0 viewsOpenAI Adds Computer-Use Agents and 20+ Plugins to ChatGPT for Healthcare
- Tech & AI x.com 1 viewGoogle warns: "the era of autonomous attacks has arrived"

