Instruction Following on ALFRED (Specific Success Rates)
91Average Task Success RateHuman Performance
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Human Performance2026.04 | 91 | — | — | — | — | — | — | |
| EUEABase Model=InternVL3-8B, Training Stage=GRPO2026.04 | 85.78 | 90.74 | 85.37 | 85.19 | 74.49 | 98.41 | 87.18 | |
| EUEABase Model=InternVL3-8B, Training Stage=SFT2026.04 | 83.45 | 90.74 | 86.59 | 75.93 | 65.31 | 98.41 | 91.03 | |
| BCBase Model=InternVL3-8B2026.04 | 74.59 | 88.89 | 73.17 | 57.41 | 62.24 | 96.83 | 75.64 | |
| EMMA2026.04 | 67.83 | 66.67 | 71.95 | 75.93 | 65.31 | 55.56 | 71.8 |