# ForceHarness: Frozen GPT-6 Astra Gets Force Control Without Fine-Tuning

> Source: <https://explainx.ai/blog/forceharness-frozen-frontier-models-force-robot-harness-2026?utm_source=rss&utm_medium=feed&utm_campaign=blog>
> Published: 2026-11-18 00:00:00+00:00

**Key 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.

**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).

If 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.

## What problem ForceHarness targets

Frontier 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):

1. **Contact evolves faster than an API model chooses actions** , so something must regulate force between model calls.
2. **Frozen models do not interpret high-rate force streams** , so raw wrenches need translation into quantities models already reason about.

That 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.

## Architecture: command path vs force path

ForceHarness diagram connecting frozen GPT-6 Astra to force primitives and physics translation for contact-rich robot manipulation
*Image: [Show Lab, National University of Singapore](https://showlab.github.io/ForceHarness/?ref=explainx)*

The 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:

### Self-terminating force primitives

Twelve 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.

### Physics translation layer

Fixed 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.

### Force sensing without a wrist F/T sensor

With 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.

## Real-robot results (Franka FR3, GPT-6 Astra)

The 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.

Those 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.

## Simulation: ForceBench and HiddenPhys

Sim-to-real transfer illustration for ForceHarness benchmarks in Isaac Lab and HiddenPhys MuJoCo

### ForceBench (Isaac Lab, RGB only)

On **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).

### HiddenPhys (MuJoCo, hidden physical variable)

**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.

They 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.

## How to try it (reproduce before you trust it)

The repo is MIT-licensed tooling around paid API calls — budget accordingly.

```
git clone https://github.com/showlab/ForceHarness
cd ForceHarness && pip install -e ".[sim]"
python scripts/check_sim.py
python scripts/run_episode.py --task weight_sort \
  --model gpt-6-astra --paper-defaults --seed 9100
```

For 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.

Paid API costs scale with episodes; start in **HiddenPhys simulation** before pointing at a factory arm.

## Why this matters if you build agentic robotics

| Design choice | ForceHarness implication | 
|---|---|
| Frozen API model + tool loop | You can add **new contact skills as tools** without retraining VLA weights. | 
| Datacenter inference | Accept **latency** ; primitives own the millisecond loop. | 
| Dataset collection | Beats **200-demo** force VLAs on public suites in their tables — shifts ROI toward**harness engineering** . | 
| 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. | 

ForceHarness 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.

## Relation to GPT-6 Astra robot claims

Several 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.

## Limitations the authors acknowledge

- **API cost and latency** cap episode length; primitives exist partly to reduce call frequency during contact.
- **Simulation seeds used during ForceBench tool development** are not held out; the page labels which rows are strictly held out.
- **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.

## Comparison snapshot

| System | Training data | Force path | Highlight | 
|---|---|---|---|
| ForceHarness + GPT-6 Astra | None (zero-shot API) | Primitives + physics layer | 54/60 real arm successes reported | 
| π₀.₅ + Prompt Fusion (ForceBench cited row) | 200 demos/task | Learned | 35.3% mean in authors’ comparison table | 
| ScaleVLA 552B VLA | Robot demos + huge VLM | Not force-first | Sample efficiency on LIBERO-PRO | 
| Expert force controller (HiddenPhys) | Hand-written FSM | Classical | 136/140, benchmark ceiling | 

## Governance and export context (why timing matters)

This 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.

## What to watch next

- Independent labs replaying **held-out seeds** on ForceBench.
- Whether **GPT-6.1 Sol + ForceHarness** closes the small gap behind Astra on the real arm table.
- 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.

## Related reading

## Episode monitoring and debugging

The [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.

Debugging checklist we extracted from the release README and paper:

1. **Confirm sim health** with`python scripts/check_sim.py` before spending API credits.
2. **Replay a known-good seed** from the paper defaults before random exploration.
3. **Compare damage counters** on HiddenPhys — trained VLAs in their table show more broken handles than ForceHarness rows.
4. **Gate hardware** behind the shipped setup skill so calibration completes before the arm moves.

## FAQ for procurement teams

Robotics 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.

## Paper metadata

| Field | Value | 
|---|---|
| arXiv | [2611.21099](https://arxiv.org/abs/2611.21099?ref=explainx) | 
| Authors | Ke Sun, Kevin Yuchen Ma, Yike Zhang, Yanzhe Chen, Mike Zheng Shou (Show Lab, NUS) | 
| Code | [github.com/showlab/ForceHarness](https://github.com/showlab/ForceHarness?ref=explainx) | 
| Data | [huggingface.co/datasets/Jack97/ForceHarness-Data](https://huggingface.co/datasets/Jack97/ForceHarness-Data?ref=explainx) | 

ForceHarness 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.
