List your company

Role-playing Agent Evaluation on PersonaGym

4.13Action JustificationGPT-4.1
3.353.55253.7553.9575May 16, 2026

Evaluation results

MethodLinks
GPT-4.12026.054.134.1344.254.884.28
DPO-Qwen3-8Btraining=Direct Preference Optimization (DPO)2026.053.883.633.754.254.924.09
SFT-Qwen3-8Btraining=Supervised Fine-Tuning (SFT)2026.053.53.633.53.884.933.88
Qwen3-8Bstatus=base model2026.053.383.133.133.754.913.66