Active Perception on Croissant Search Active Perception
24.3STCoMe-VLA
Evaluation results
| Method | Links | ||
|---|---|---|---|
| CoMe-VLAData Composition=800k human + 400k robot samples2026.02 | 24.3 | — | |
| CoMe-VLAData Composition=800k human + 100k robot samples2026.02 | 26.7 | — | |
| CoMe-VLAData Composition=400k human + 400k robot samples2026.02 | 28.4 | — | |
| π0.52026.02 | 116.6 | — | |
| CoMe-VLAData Composition=0 human + 400k robot samples2026.02 | 237.8 | — | |
| DP2026.02 | 431.8 | — | |
| OpenVLA-OFT2026.02 | 540.7 | — | |
| ACT2026.02 | 591.9 | — |