Social Deduction Game Gameplay on ONUW
61.7Village Team Win RateOurs + ReAct
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Ours + ReActBackend LLM=Gemini-2.5-Flash2025.10 | 61.7 | 39.6 | 45.1 | |
| ReActBackend LLM=Gemini-2.5-Flash2025.10 | 56.1 | 40 | 43.1 | |
| Ours + RL-ins.Backend LLM=Gemini-2.5-Flash2025.10 | 55.2 | 50.6 | 51.5 | |
| BeliefBackend LLM=Gemini-2.5-Flash2025.10 | 54.6 | 40.9 | 43.5 | |
| RL-ins.Backend LLM=Gemini-2.5-Flash2025.10 | 54.2 | 47.4 | 48.5 | |
| LLM-ins.Backend LLM=Gemini-2.5-Flash2025.10 | 54.1 | 43.9 | 45.9 |