Tactile Reasoning and Understanding on PHYSICLEAR (test)
57.3PCUniTac-7B
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| UniTac-7BType=Und. and Gen., Protocol=AR + Diff2026.06 | 57.3 | 64.61 | 59.22 | 61.53 | 74.35 | 82.05 | 66.51 | |
| UniTac-3BType=Und. and Gen., Protocol=AR + Diff2026.06 | 54.97 | 42.13 | 58.9 | 51.28 | 76.92 | 79.48 | 60.61 | |
| Octopi-7BType=Und. Only, Protocol=AR2026.06 | 45.5 | 22.22 | 48 | 64.1 | 73.92 | 87.17 | 57.31 | |
| TokenFlow-7BType=Und. and Gen., Protocol=AR2026.06 | 38.05 | 29.48 | 32.56 | 47.68 | 38.7 | 48.31 | 39.13 | |
| LLaVA-OneVision-7BType=Und. Only, Protocol=AR2026.06 | 35.42 | 29.11 | 28.44 | 33.33 | 32.18 | 36.51 | 32.49 | |
| GPT-4oType=Und. Only, Protocol=AR2026.06 | 30.87 | 22.62 | 38.05 | 31.37 | 30.48 | 36.51 | 31.65 | |
| JanusPro-7BType=Und. and Gen., Protocol=AR + Diff2026.06 | 30.18 | 26.71 | 38.05 | 41.63 | 35.82 | 45.37 | 36.29 | |
| Gemini-2.5-Pro-ExpType=Und. Only, Protocol=AR2026.06 | 26.71 | 23.47 | 25.06 | 29.19 | 28.44 | 32.18 | 27.5 | |
| BLIP3o-3BType=Und. and Gen., Protocol=AR + Diff2026.06 | 26.71 | 21.47 | 32.18 | 36.51 | 32.18 | 48 | 32.84 | |
| Qwen2.5-VL-7BType=Und. Only, Protocol=AR2026.06 | 21.47 | 19.92 | 23.05 | 25.73 | 33.33 | 26.62 | 25.01 |