Embodied Task Execution on VLABench Avg.
65.9Progress Score (PS)ERVLA
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| ERVLAmode=Zero-shot2026.06 | 65.9 | 53.2 | 70.4 | |
| π0.5mode=Zero-shot2026.06 | 62.3 | 48.1 | 64.9 | |
| Choice + No Knowledge TruncationAblation=Choice without KT, mode=Zero-shot2026.06 | 59.8 | 47.2 | 63.4 | |
| No Choice + Knowledge InsulationAblation=No choice and KI, mode=Zero-shot2026.06 | 52.5 | 42.6 | 58.4 | |
| X-VLAmode=Zero-shot2026.06 | 51.1 | — | — | |
| No CoTAblation=No explicit chain-of-thought, mode=Zero-shot2026.06 | 50.4 | 40.9 | 57.1 | |
| π0-FASTmode=Zero-shot2026.06 | 49.5 | 39.8 | 58.6 | |
| No Choice (End-to-End)Ablation=End-to-End No Choice, mode=Zero-shot2026.06 | 48.2 | 39.2 | 55.8 | |
| ACoT-VLAmode=Zero-shot2026.06 | 47.4 | — | 63.5 | |
| π0mode=Zero-shot2026.06 | 44.1 | 29.4 | 55 |