Code Generation on MBPP (MBPP Score)
35.6MBPP ScoreWIDEN
Evaluation results
| Method | Links | |
|---|---|---|
| WIDENExpert Combination=Code + Instruction2026.04 | 35.6 | |
| DARE TiesExpert Combination=Code + Instruction2026.04 | 35.2 | |
| Precise FusionExpert Combination=Code + Instruction2026.04 | 35.2 | |
| Ties MergingExpert Combination=Instruction + Math2026.04 | 34.6 | |
| DARE Task ArithmeticExpert Combination=Code + Instruction2026.04 | 34.4 | |
| Task ArithmeticExpert Combination=Code + Instruction2026.04 | 33.8 | |
| WIDENExpert Combination=Code + Instruction + Math2026.04 | 33.2 | |
| Precise FusionExpert Combination=Instruction + Math2026.04 | 32.4 | |
| Instruction TunedModel Category=Base Models2026.04 | 32.2 | |
| Avg BaselineExpert Combination=Code + Instruction2026.04 | 31.6 | |
| Precise FusionExpert Combination=Code + Instruction + Math2026.04 | 31 | |
| Precise FusionExpert Combination=Code + Math2026.04 | 30.6 | |
| WIDENExpert Combination=Instruction + Math2026.04 | 30.4 | |
| Avg BaselineExpert Combination=Code + Math2026.04 | 30.2 | |
| Avg BaselineExpert Combination=Code + Instruction + Math2026.04 | 30.2 | |
| WIDENExpert Combination=Code + Math2026.04 | 29.4 | |
| CodeModel Category=Base Models2026.04 | 29.2 | |
| Ties MergingExpert Combination=Code + Instruction + Math2026.04 | 29.2 | |
| Avg BaselineExpert Combination=Instruction + Math2026.04 | 29 | |
| DARE Task ArithmeticExpert Combination=Code + Math2026.04 | 28.6 | |
| Task ArithmeticExpert Combination=Code + Math2026.04 | 28.6 | |
| BaseModel Category=Base Models2026.04 | 28.2 | |
| DARE TiesExpert Combination=Code + Math2026.04 | 27.4 | |
| Ties MergingExpert Combination=Code + Math2026.04 | 26.8 | |
| DARE TiesExpert Combination=Instruction + Math2026.04 | 23.8 | |
| Ties MergingExpert Combination=Code + Instruction2026.04 | 23.6 | |
| MathModel Category=Base Models2026.04 | 23.4 | |
| DARE TiesExpert Combination=Code + Instruction + Math2026.04 | 21.2 | |
| Task ArithmeticExpert Combination=Instruction + Math2026.04 | 20.2 | |
| DARE Task ArithmeticExpert Combination=Code + Instruction + Math2026.04 | 19.6 | |
| Task ArithmeticExpert Combination=Code + Instruction + Math2026.04 | 19.6 | |
| DARE Task ArithmeticExpert Combination=Instruction + Math2026.04 | 19 |