Social agent strategy optimization on Sotopia subset
7.85Goal ScoreThink-on-Your-Feet (AMPO)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Think-on-Your-Feet (AMPO)Evaluation setting=Qwen-7B-Instruct self-play2026.05 | 7.85 | 3.54 | |
| Sotopia-Ω (DSI)Evaluation setting=Qwen-7B-Instruct self-play2026.05 | 7.31 | 3.51 | |
| ALSOEvaluation setting=Qwen-7B-Instruct self-play2026.05 | 7.26 | 3.58 |