Action-Conditioned Hand-Object Co-Grounded Reasoning on Know-Show (test)
100T/AccHuman
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| HumanSubset=Know-Show subset2025.12 | 100 | 83.6 | 85.7 | 91.1 | 79.5 | |
| Gemini-2.5-ProSubset=Know-Show subset2025.12 | 32 | 4 | 8 | 12 | 0 | |
| Gemini-2.5-FlashSubset=Know-Show subset2025.12 | 8 | 0 | 4 | 12 | 0 | |
| GRAM2025.12 | 4.2 | 1 | 0.5 | 0 | 0 | |
| GPT-4oSubset=Know-Show subset2025.12 | 4 | 0 | 0 | 0 | 0 | |
| LLaVA-OneVision2025.12 | 3.7 | 0.8 | 0.5 | 0.8 | 0 | |
| Ovis2.5-9B2025.12 | 3.5 | 0.3 | 0.3 | 0.3 | 0 | |
| VideoLLaVA2025.12 | 3.2 | 0.8 | 0.8 | 0 | 0 | |
| MiniCPM-V 4.52025.12 | 3.2 | 1.2 | 0.8 | 0.3 | 0 | |
| Qwen2.5 VL 7B2025.12 | 2.5 | 3.5 | 2 | 1.5 | 0 |