Social Skill on Sotopia
47Primary MetricDITTO
Evaluation results
| Method | Links | |
|---|---|---|
| DITTOBackbone=Qwen3-VL-8B-Instruct2026.05 | 47 | |
| GRPOBackbone=Qwen3-VL-8B-Instruct2026.05 | 42.3 | |
| Sotopia-RL-7B2026.05 | 31.2 | |
| GPT-5-nano2026.05 | 31 | |
| GPT-5.42026.05 | 30 | |
| Qwen3-VL-8B-InstructRole=Base2026.05 | 27.7 |