Medical Question Answering on HealthBench Medicine N=5,000 (overall)
26.1Rubric ScoreRVPO (k=0.5→2.0 / 1.0→2.0)
Evaluation Results
| Method | Links | |
|---|---|---|
| RVPO (k=0.5→2.0 / 1.0→2.0)Size=14B, Checkpoint=Best Ckpt, k=0.5→2.0 / 1.0→2.02026.05 | 26.1 | |
| GRPO (Explicit)Size=14B, Checkpoint=Best Ckpt2026.05 | 23.6 | |
| RVPO (k=0.5→2.0 / 1.0→2.0)Size=14B, Checkpoint=Final (300), k=0.5→2.0 / 1.0→2.02026.05 | 23.6 | |
| GRPO (Implicit)Size=14B, Checkpoint=Best Ckpt2026.05 | 23.4 | |
| GRPO (Implicit)Size=14B, Checkpoint=Final (300)2026.05 | 23.4 | |
| RVPO (k=0.5→2.0 / 1.0→2.0)Size=7B, Checkpoint=Best Ckpt, k=0.5→2.0 / 1.0→2.02026.05 | 23 | |
| RVPO-explicit (β=1.0/0.5)Size=7B, Checkpoint=Best Ckpt, β=1.0/0.52026.05 | 22.7 | |
| RVPO-min (k=∞)Size=14B, Checkpoint=Best Ckpt, k=∞2026.05 | 22.5 | |
| GRPO (Explicit)Size=7B, Checkpoint=Best Ckpt2026.05 | 22.1 | |
| GDPOSize=14B, Checkpoint=Best Ckpt2026.05 | 21.5 | |
| RVPO (k=0.5→2.0 / 1.0→2.0)Size=7B, Checkpoint=Final (300), k=0.5→2.0 / 1.0→2.02026.05 | 20.4 | |
| GDPOSize=7B, Checkpoint=Best Ckpt2026.05 | 19.8 | |
| GRPO (Implicit)Size=7B, Checkpoint=Best Ckpt2026.05 | 19.3 | |
| GRPO (Implicit)Size=7B, Checkpoint=Final (300)2026.05 | 19.3 | |
| GDPOSize=3B, Checkpoint=Best Ckpt2026.05 | 19.2 | |
| RVPO-min (k=∞)Size=7B, Checkpoint=Best Ckpt, k=∞2026.05 | 19.1 | |
| GRPO (Explicit)Size=3B, Checkpoint=Best Ckpt2026.05 | 19 | |
| RVPO-min (k=∞)Size=14B, Checkpoint=Final (300), k=∞2026.05 | 19 | |
| RVPO (k=0.5→2.0 / 1.0→2.0)Size=3B, Checkpoint=Best Ckpt, k=0.5→2.0 / 1.0→2.02026.05 | 18.9 | |
| RVPO-explicit (β=1.0/0.5)Size=14B, Checkpoint=Best Ckpt, β=1.0/0.52026.05 | 18.8 | |
| RVPO-explicit (β=1.0/0.5)Size=3B, Checkpoint=Best Ckpt, β=1.0/0.52026.05 | 18.4 | |
| RVPO-min (k=∞)Size=3B, Checkpoint=Best Ckpt, k=∞2026.05 | 18.1 | |
| RVPO-explicit (β=1.0/0.5)Size=7B, Checkpoint=Final (300), β=1.0/0.52026.05 | 17.8 | |
| RVPO-explicit (β=1.0/0.5)Size=14B, Checkpoint=Final (300), β=1.0/0.52026.05 | 16.3 | |
| GRPO (Implicit)Size=3B, Checkpoint=Best Ckpt2026.05 | 15.4 | |
| RVPO (k=0.5→2.0 / 1.0→2.0)Size=3B, Checkpoint=Final (300), k=0.5→2.0 / 1.0→2.02026.05 | 14.7 | |
| RVPO-min (k=∞)Size=3B, Checkpoint=Final (300), k=∞2026.05 | 12.4 | |
| GDPOSize=3B, Checkpoint=Final (300)2026.05 | 11.7 | |
| GRPO (Explicit)Size=7B, Checkpoint=Final (300)2026.05 | 10.2 | |
| GRPO (Implicit)Size=3B, Checkpoint=Final (300)2026.05 | 7.2 | |
| GRPO (Explicit)Size=3B, Checkpoint=Final (300)2026.05 | 5.3 | |
| GDPOSize=7B, Checkpoint=Final (300)2026.05 | 2.6 | |
| RVPO-explicit (β=1.0/0.5)Size=3B, Checkpoint=Final (300), β=1.0/0.52026.05 | 1.1 | |
| RVPO-min (k=∞)Size=7B, Checkpoint=Final (300), k=∞2026.05 | 0 | |
| GRPO (Explicit)Size=14B, Checkpoint=Final (300)2026.05 | 0 | |
| GDPOSize=14B, Checkpoint=Final (300)2026.05 | 0 |