ATOM-Bench - Dual-Arm Instruction: leaderboard
The dual-arm Instruction Set of ATOM-Bench (Wu et al., 2026) on a real Agilex Cobot Magic robot: seven bimanual tasks that each isolate one language constraint (color, shape, size, spatial relation, count, destination, exclusion). Success rate in percent over 10 fixed physical test seeds per task after fine-tuning on 100 human demonstrations per atomic task, macro-averaged; higher is better.
Metric: Success rate (%; Instruction Set, 7 tasks x 10 fixed physical test seeds on the Agilex Cobot Magic dual-arm track, macro-averaged over tasks; one policy fine-tuned on the 15 atomic tasks of the track). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Pi0.5 | 71.4 | |
| 2 | Motus | 50 | |
| 3 | GR00T-N1.6 | 27.1 | |
| 4 | LingBot-VLA | 21.4 | |
| 5 | SmolVLA | 5.7 |
Interactive version: theaggregate.ai/benchmark?slug=atom-bench-dual-arm-instruction · How It Works · Data refreshed daily, snapshot 2026-11-12.