Code Generation on Python (held-out eval)
37.7Pass@1Qwen2.5-32B-Instruct (frozen)
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen2.5-32B-Instruct (frozen)Configuration=Qwen2.5-32B-Instruct (frozen), Training=Frozen, Seed count=n=22026.06 | 37.7 | |
| Qwen2.5-7B baseConfiguration=Qwen2.5-7B base, Training=No training, Seed count=n=12026.06 | 25 | |
| Qwen2.5-7B + CARE 5 campaignsConfiguration=Qwen2.5-7B + CARE 5 campaigns, Training=5-campaign training, Seed count=n=32026.06 | 20 | |
| Qwen2.5-7B + naive 5 campaignsConfiguration=Qwen2.5-7B + naive 5 campaigns, Training=5-campaign training, Seed count=n=32026.06 | 12.3 |