Compositional Information Seeking on CodeClinic (test)
57.1Level 1 AccuracyGRPO Finetuning
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| GRPO FinetuningMode=GRPO Finetuning2026.05 | 57.1 | 36.5 | 27.9 | 40.5 | 22,376 | 2.7 | |
| Clinical AutoformalizationMode=Autoformalization2026.05 | 51.2 | 41.5 | 34.1 | 45.3 | 23,569 | 2.5 | |
| ZeroshotMode=Zeroshot2026.05 | 44.9 | 32.8 | 26.3 | 38.1 | 30,479 | 3.1 |