Medical Differential Diagnoses on DDXPlus
79Avg CorrectMulti-Answer RLVR
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Multi-Answer RLVRLoss type=Multi Loss, Prompting strategy=Multi Prompt, k-value=32026.03 | 79 | 100 | 677 | 42 | |
| Multi-Answer RLCRLoss type=Multi Loss, Prompting strategy=Multi Prompt, k-value=32026.03 | 77 | 100 | 510 | 43 | |
| RLVR (Single Loss + Multi Prompt)Loss type=Single Loss, Prompting strategy=Multi Prompt, k-value=32026.03 | 67 | 99 | 695 | 42 | |
| RLCR (Single Loss + Single Prompt)Loss type=Single Loss, Prompting strategy=Single Prompt, k-value=32026.03 | 65 | 50 | 1,378 | 55 | |
| RLVR (Single Loss + Single Prompt)Loss type=Single Loss, Prompting strategy=Single Prompt, k-value=32026.03 | 62 | 62 | 1,191 | 50 | |
| RLCR (Single Loss + Multi Prompt)Loss type=Single Loss, Prompting strategy=Multi Prompt, k-value=32026.03 | 49 | 100 | 703 | 32 | |
| Zero-Shot (Multi Prompt)Prompting strategy=Multi Prompt, k-value=3, Comparison Baseline=RLCR group2026.03 | 41 | 89 | 1,269 | 16 | |
| Zero-Shot (Multi Prompt)Prompting strategy=Multi Prompt, k-value=3, Comparison Baseline=RLVR group2026.03 | 31 | 93 | 1,131 | 24 |