Empathy assessment on EMPA
27Pass RateGemini-2.5-pro
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Gemini-2.5-proModel=Gemini-2.5-pro2026.03 | 27 | 90.7 | |
| MAPOModel=Qwen3-32B2026.03 | 26 | 84.3 | |
| Claude-3.5-sonnetModel=Claude-3.5-sonnet2026.03 | 25 | 85.1 | |
| DeepSeek-V3.2Model=DeepSeek-V3.22026.03 | 25 | 78.4 | |
| GRPOModel=Qwen3-32B2026.03 | 22 | 73.9 | |
| MAPOModel=Qwen3-14B2026.03 | 20 | 67.8 | |
| BaseModel=Qwen3-32B2026.03 | 19 | 68.9 | |
| BaseModel=Qwen3-14B2026.03 | 12 | 53.5 | |
| GRPOModel=Qwen3-14B2026.03 | 12 | 55.4 | |
| MAPOModel=Qwen2.5-7B-instruct2026.03 | 9 | 58.9 | |
| MAPOModel=Qwen3-8B2026.03 | 8 | 41.6 | |
| GRPOModel=Qwen3-8B2026.03 | 5 | 33 | |
| GRPOModel=Qwen2.5-7B-instruct2026.03 | 1 | 28 | |
| BaseModel=Qwen3-8B2026.03 | 0 | 13.3 | |
| BaseModel=Qwen2.5-7B-instruct2026.03 | 0 | 15.7 |