{"slug": "forceharness-frozen-gpt-6-astra-gets-force-control-without-fine-tuning", "title": "ForceHarness: Frozen GPT-6 Astra Gets Force Control Without Fine-Tuning", "summary": "Show Lab at the National University of Singapore released ForceHarness, an open agent harness that lets frozen frontier API models perform contact-rich robot manipulation without fine-tuning or human demonstrations. On a Franka FR3 arm with joint torques but no wrist force sensor, the authors report GPT-6 Astra rising from 2/60 to 54/60 successful episodes across six tasks, summarized as 3.3% to 90.0% zero-shot success. The harness runs self-terminating force primitives locally beneath the model and converts force traces into mass, friction, contact height and stiffness estimates with uncertainty windows, addressing the gap where contact evolves faster than an API model chooses actions.", "body_md": "**Show Lab at the National University of Singapore released [ForceHarness](https://showlab.github.io/ForceHarness/?ref=explainx), an open agent harness that lets frozen frontier API models do contact-rich manipulation by closing a fast force loop beneath the model and returning physics the model can reason about — with no finetuning and no human demonstrations.** On a real **Franka FR3** arm that had **joint torques but no wrist force sensor**, the authors report **GPT-6 Astra** rising from **2/60** to **54/60** successful episodes across six tasks, and they summarize that as **3.3% → 90.0%** zero-shot success in the [paper abstract](https://arxiv.org/abs/2611.21099?ref=explainx).\n\nIf you already ship [vision-language-action stacks](https://explainx.ai/blog/scalevla-552b-vision-language-action-robot-backbone-2026) or [GR00T-style sim glue](https://explainx.ai/blog/nvidia-simready-central-pick-and-place-isaac-sim-6-2-gpt-6-sol-agent-ur17g-robotiq-2026), the pitch is narrow but sharp: **RGB-only policies still fail when the task is decided by touch** — whether a USB connector seated, a screw caught thread, or one bottle is heavier than another. ForceHarness is the missing **harness layer** between a frozen multimodal model and a robot arm that already has compliant control.\n\n## What problem ForceHarness targets\n\nFrontier models can plan pick-and-place from images. Contact-rich skills break that assumption for two reasons the authors spell out on the [project page](https://showlab.github.io/ForceHarness/?ref=explainx):\n\n1. **Contact evolves faster than an API model chooses actions** , so something must regulate force between model calls.\n2. **Frozen models do not interpret high-rate force streams** , so raw wrenches need translation into quantities models already reason about.\n\nThat is the same gap [CoRL 2026 career advice](https://explainx.ai/blog/yann-lecun-corl-2026-dont-work-on-llms-jepa-career-advice) keeps highlighting: language-scale models and **touch-scale physics** run on different clocks.\n\n## Architecture: command path vs force path\n\nForceHarness diagram connecting frozen GPT-6 Astra to force primitives and physics translation for contact-rich robot manipulation\n*Image: [Show Lab, National University of Singapore](https://showlab.github.io/ForceHarness/?ref=explainx)*\n\nThe harness splits into a **blue command path** (images in, tool calls out) and a **teal force path** (primitives run locally, return structured physics). Three pieces matter for builders:\n\n### Self-terminating force primitives\n\nTwelve contact skills (`search`, `wiggle`, `screw`, `slide`, `lift`, and others) share one control loop. A **target pose leads measured pose** so the arm’s impedance controller turns lead into force; an admittance law holds commanded force; the primitive **stops on contact, stall, resistance drop, or its internal goal**, then holds still while the model thinks. That is how the system respects API latency without dropping torque safety.\n\n### Physics translation layer\n\nFixed estimators turn each primitive’s trace into **mass, centre-of-mass offset, friction, contact height, stiffness, insertion advance** — each with **uncertainty windows**, no hand-tuned task thresholds. The layer reports **how heavy** a load is, not **whether that is heavy enough**, so the frontier model makes the judgment call.\n\n### Force sensing without a wrist F/T sensor\n\nWith a wrist sensor, hand force is direct. Without one, the stack solves **w = (Jᵀ)⁺ τ_ext** from joint torques. Three runtime rules handle bias: **tare at rest**, **settle then read**, and **report deltas** rather than absolutes — important on arms like the FR3 setup in the paper.\n\n## Real-robot results (Franka FR3, GPT-6 Astra)\n\nThe project page publishes a **per-task success table** (10 episodes each) for six contact-rich skills: USB insertion, screw driving, eggplant peeling, whiteboard wiping, bottle weight sorting, and firmness probing. **No-force GPT-6 Astra** totals **2/60**; **ForceHarness** totals **54/60**; **GPT-6.1 Sol + ForceHarness** lands **47/60**. The authors note **Fisher exact p < 0.01** between force and no-force conditions for GPT-6 Astra.\n\nThose tasks are deliberately **force-first**: weight sort and firmness probe need sensing the camera cannot see. That aligns with [Hubego’s contact-rich data pitch](https://explainx.ai/blog/hubego-100k-skilled-labor-video-open-robot-training-2026) — internet video rarely shows fine contact.\n\n## Simulation: ForceBench and HiddenPhys\n\nSim-to-real transfer illustration for ForceHarness benchmarks in Isaac Lab and HiddenPhys MuJoCo\n\n### ForceBench (Isaac Lab, RGB only)\n\nOn **held-out seeds** the release had not used before, **GPT-6 Astra + ForceHarness** averages **82.5%** across eight ForceBench tasks versus **62.5%** without force and **35.3%** for the best **trained** row reported in the original ForceBench paper (200 demos per task, not reproduced by the authors). Details and seed splits are on the [project page tables](https://showlab.github.io/ForceHarness/?ref=explainx).\n\n### HiddenPhys (MuJoCo, hidden physical variable)\n\n**HiddenPhys** is their new seven-task suite where success requires inferring a physical property the image hides. **GPT-6 Astra + ForceHarness** scores **135/140** total successes with **0 damaged handles** in their table, matching a hand-written expert force controller (**136/140**, 1 damage) and beating demonstration-trained **π₀.₅**, **ForceVLA**, and **Diffusion Policy** rows that show higher damage counts.\n\nThey ship **1,423 expert episodes** on [Hugging Face dataset Jack97/ForceHarness-Data](https://huggingface.co/datasets/Jack97/ForceHarness-Data?ref=explainx) for replay and LeRobot training.\n\n## How to try it (reproduce before you trust it)\n\nThe repo is MIT-licensed tooling around paid API calls — budget accordingly.\n\n```\ngit clone https://github.com/showlab/ForceHarness\ncd ForceHarness && pip install -e \".[sim]\"\npython scripts/check_sim.py\npython scripts/run_episode.py --task weight_sort \\\n  --model gpt-6-astra --paper-defaults --seed 9100\n```\n\nFor agent-assisted setup, the authors ship **`skills/forceharness-setup/SKILL.md`**: paste their prompt into **Claude Code, Codex, or Cursor** so the agent installs simulators gate-by-gate and **never moves hardware without confirmation**. That pattern mirrors how explainx.ai covers [external sandboxes for coding agents](https://explainx.ai/blog/external-sandboxes-coding-agents-prime-radiant-willison-november-2026) — safety gates belong in the harness, not the model card.\n\nPaid API costs scale with episodes; start in **HiddenPhys simulation** before pointing at a factory arm.\n\n## Why this matters if you build agentic robotics\n\n| Design choice | ForceHarness implication | \n|---|---|\n| Frozen API model + tool loop | You can add **new contact skills as tools** without retraining VLA weights. | \n| Datacenter inference | Accept **latency** ; primitives own the millisecond loop. | \n| Dataset collection | Beats **200-demo** force VLAs on public suites in their tables — shifts ROI toward**harness engineering** . | \n| Safety | Damage counts on HiddenPhys are explicit; pair with [ViCoNE physical-AI CVE work](https://explainx.ai/blog/vicone-28-zero-day-robotics-vulnerabilities-physical-ai-ieee-2026) for deployment hardening. | \n\nForceHarness does **not** replace [NVIDIA GR00T](https://explainx.ai/blog/agentic-robotics-berkeley-nvidia-corl-2026) or [UniRobot](https://explainx.ai/blog/uni-robot-agent-open-qwen-robodojo) training stacks. It argues a **complementary layer**: when the model is already frontier-scale and API-only, **force tools + physics summaries** may outperform another fine-tune.\n\n## Relation to GPT-6 Astra robot claims\n\nSeveral 2026 robot posts cite **GPT-6 Astra** manipulation demos. ForceHarness is the first open release we saw that publishes **full episode tables**, **held-out seeds**, and **damage metrics** side by side with **no-force** baselines on identical harness settings. Treat marketing clips and paper tables differently until you replay seeds locally.\n\n## Limitations the authors acknowledge\n\n- **API cost and latency** cap episode length; primitives exist partly to reduce call frequency during contact.\n- **Simulation seeds used during ForceBench tool development** are not held out; the page labels which rows are strictly held out.\n- **Smaller models** still struggle on some tasks even with physics translation — read ablations in[arXiv:2611.21099](https://arxiv.org/abs/2611.21099?ref=explainx) before swapping models.\n\n## Comparison snapshot\n\n| System | Training data | Force path | Highlight | \n|---|---|---|---|\n| ForceHarness + GPT-6 Astra | None (zero-shot API) | Primitives + physics layer | 54/60 real arm successes reported | \n| π₀.₅ + Prompt Fusion (ForceBench cited row) | 200 demos/task | Learned | 35.3% mean in authors’ comparison table | \n| ScaleVLA 552B VLA | Robot demos + huge VLM | Not force-first | Sample efficiency on LIBERO-PRO | \n| Expert force controller (HiddenPhys) | Hand-written FSM | Classical | 136/140, benchmark ceiling | \n\n## Governance and export context (why timing matters)\n\nThis drop landed during **[Nvidia Megaspeed export-control headlines](https://explainx.ai/blog/nvidia-huang-megaspeed-alice-huang-full-story-timeline-what-happened-2026)** and **[Waymo regulatory resets](https://explainx.ai/blog/waymo-california-dmv-freeway-ban-november-2026)**. None of that changes how you `pip install` ForceHarness, but teams shipping **embodied agents in regulated markets** should keep **hardware procurement** separate from **model API geography** — the harness runs wherever your arm and inference endpoint live.\n\n## What to watch next\n\n- Independent labs replaying **held-out seeds** on ForceBench.\n- Whether **GPT-6.1 Sol + ForceHarness** closes the small gap behind Astra on the real arm table.\n- Integration patterns with [MCP tool hosts](https://explainx.ai/blog/mcp-tool-provider-misbinding-ai-agents-security-2026) so force tools sit beside camera tools in one agent schema.\n\n## Related reading\n\n## Episode monitoring and debugging\n\nThe [project page](https://showlab.github.io/ForceHarness/?ref=explainx) describes a **web monitor** that runs beside real-robot episodes: live wrist force, model observations, tool calls, and structured tool results, with recordings for side-by-side review. That matters because contact failures are often **one wrong primitive** — not a bad high-level plan. If you already run [coding-agent external sandboxes](https://explainx.ai/blog/external-sandboxes-coding-agents-prime-radiant-willison-november-2026), treat force episodes the same way: **log every tool boundary** before you blame the base model.\n\nDebugging checklist we extracted from the release README and paper:\n\n1. **Confirm sim health** with`python scripts/check_sim.py` before spending API credits.\n2. **Replay a known-good seed** from the paper defaults before random exploration.\n3. **Compare damage counters** on HiddenPhys — trained VLAs in their table show more broken handles than ForceHarness rows.\n4. **Gate hardware** behind the shipped setup skill so calibration completes before the arm moves.\n\n## FAQ for procurement teams\n\nRobotics leads asking vendors “Do you support force-aware API agents?” can cite ForceHarness as an **open reference architecture**, not a warranty. It does not certify ISO safety; it shows how **tool schemas** can wrap **admittance control** under **frozen weights**. Pair vendor answers with [Waymo’s regulatory story](https://explainx.ai/blog/waymo-california-dmv-freeway-ban-november-2026): regulators care about **failure modes**, not benchmark slides.\n\n## Paper metadata\n\n| Field | Value | \n|---|---|\n| arXiv | [2611.21099](https://arxiv.org/abs/2611.21099?ref=explainx) | \n| Authors | Ke Sun, Kevin Yuchen Ma, Yike Zhang, Yanzhe Chen, Mike Zheng Shou (Show Lab, NUS) | \n| Code | [github.com/showlab/ForceHarness](https://github.com/showlab/ForceHarness?ref=explainx) | \n| Data | [huggingface.co/datasets/Jack97/ForceHarness-Data](https://huggingface.co/datasets/Jack97/ForceHarness-Data?ref=explainx) | \n\nForceHarness is **education-grade open tooling**: read the paper, run HiddenPhys, and only then decide whether your product needs another VLA fine-tune or a **force-aware tool harness** on the model you already pay for.", "url": "https://wpnews.pro/news/forceharness-frozen-gpt-6-astra-gets-force-control-without-fine-tuning", "canonical_source": "https://explainx.ai/blog/forceharness-frozen-frontier-models-force-robot-harness-2026?utm_source=rss&utm_medium=feed&utm_campaign=blog", "published_at": "2026-11-18 00:00:00+00:00", "updated_at": "2026-11-18 05:18:33.063587+00:00", "lang": "en", "topics": ["robotics", "ai-agents", "ai-research", "artificial-intelligence", "ai-tools"], "entities": ["Show Lab", "National University of Singapore", "ForceHarness", "GPT-6 Astra", "Franka FR3", "GPT-6.1 Sol"], "also_reported_by": [], "takeaway": "Wrapping a frozen API model in a local force-control harness with 12 self-terminating primitives raised GPT-6 Astra from 2/60 to 54/60 zero-shot episodes on a Franka FR3 without a wrist force sensor, estimating wrench as w = (Jᵀ)⁺ τ_ext from joint torques.", "recommendation": null, "alternates": {"html": "https://wpnews.pro/news/forceharness-frozen-gpt-6-astra-gets-force-control-without-fine-tuning", "markdown": "https://wpnews.pro/news/forceharness-frozen-gpt-6-astra-gets-force-control-without-fine-tuning.md", "text": "https://wpnews.pro/news/forceharness-frozen-gpt-6-astra-gets-force-control-without-fine-tuning.txt", "jsonld": "https://wpnews.pro/news/forceharness-frozen-gpt-6-astra-gets-force-control-without-fine-tuning.jsonld"}}