MetaFine - Press Part: leaderboard
Fine-grained robot manipulation from MetaFine (Xu et al., 2026): the policy must press the instructed part of a switch or stapler, and a trial counts only when the part-level or directional constraint is met. Success rate in percent in simulation under nominal conditions, with every policy fine-tuned on the MetaFine demonstrations using its official recipe; higher is better.
Metric: Success rate (%; fine-grained acceptance: press the instructed part (switch, stapler head or top); nominal simulation conditions; each policy fine-tuned on the MetaFine demonstrations from its released checkpoint with the authors' official recipe). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | π0.5 | 68 | |
| 2 | π0 | 65 | |
| 3 | ACT | 63 | |
| 4 | DP3 | 63 | |
| 5 | OpenVLA-OFT | 59 | |
| 6 | OpenVLA | 53 | |
| 7 | Octo | 44 |
Interactive version: theaggregate.ai/benchmark?slug=metafine-press-part · How It Works · Data refreshed daily, snapshot 2026-11-12.