Track Prediction on CALVIN ABC → D (test)
43.7Success Rate (δ < 4)ARFM
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| ARFMProtocol=fine-tuned2025.12 | 43.7 | 57.8 | 80 | 94.4 | 98.5 | 74.9 | 19.1 | |
| No movement2025.12 | 36.7 | 39.1 | 44.2 | 53.3 | 71.9 | 49 | 37.2 | |
| ARFMProtocol=zero-shot2025.12 | 36.5 | 40.5 | 48.6 | 65.9 | 90.9 | 56.5 | 33.6 | |
| iVideoGPT2025.12 | 35.9 | 43.4 | 50.1 | 60 | 77.2 | 53 | 34.1 | |
| Cosmos Video2World2025.12 | 25.4 | 33.6 | 41.8 | 52.9 | 73 | 45.3 | 39.6 | |
| Chronos Relative2025.12 | 23.5 | 33.9 | 42.9 | 52.9 | 71.7 | 44.8 | 30.2 | |
| Chronos Absolute2025.12 | 17.2 | 23 | 31.7 | 45.1 | 66.9 | 36.8 | 84.9 |