Code Generation on BigCodeBench (BigCodeBench Metric)
24.2BigCodeBench ScoreTaskFT-SPEED-24+BoS
Evaluation Results
| Method | Links | |
|---|---|---|
| TaskFT-SPEED-24+BoSFine-tuning=OpenCodeInstruct, K=24, Visibility Policy=SPEED+BoS2026.05 | 24.2 | |
| TaskFT-SPEED-28+BoSFine-tuning=OpenCodeInstruct, K=28, Visibility Policy=SPEED+BoS2026.05 | 23.8 | |
| TaskFT-FullFine-tuning=OpenCodeInstruct, Depth=Full2026.05 | 23.7 | |
| TaskFT-SPEED-20+BoSFine-tuning=OpenCodeInstruct, K=20, Visibility Policy=SPEED+BoS2026.05 | 20.3 | |
| MONA2026.05 | 0.3377 | |
| AdamW2026.05 | 0.3123 | |
| Muon2026.05 | 0.3061 |