Multi-Modal Social Interaction on Werewolf Among Us (YouTube)
37.46STIOmni-MMSI-R
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Omni-MMSI-R2026.03 | 37.46 | 56.62 | 47.04 | |
| Gemini 2.5 Pro2026.03 | 36.13 | 53.47 | 44.8 | |
| Lee et al.*2026.03 | 29.01 | 34.8 | 31.91 | |
| Li et al.*2026.03 | 26.3 | 30.14 | 28.22 | |
| HumanOmni2026.03 | 20.8 | 28.02 | 24.41 | |
| OmniVinci2026.03 | 19.57 | 29.17 | 24.37 | |
| Qwen3 Omni 30BModel Size=30B2026.03 | 18.81 | 36.85 | 27.83 | |
| Phi-4-Multimodal2026.03 | 16.21 | 21.11 | 18.66 | |
| Qwen2.5 Omni 7BModel Size=7B2026.03 | 14 | 26.18 | 20.09 | |
| R1-Omni2026.03 | 0.15 | 0.19 | 0.17 |