Code Generation on CodeContest Overall
27.97PRDisenLoRA
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DisenLoRAFinetune Method=DisenLoRA, Base Model=Meta-llama-3.1-instruct-8b2025.02 | 27.97 | 18.24 | |
| SFT on cluster 1Finetune Method=SFT on cluster 1, Base Model=Meta-llama-3.1-instruct-8b2025.02 | 24.56 | 16.93 | |
| SFT on cluster 2Finetune Method=SFT on cluster 2, Base Model=Meta-llama-3.1-instruct-8b2025.02 | 24.38 | 16.93 | |
| TiesFinetune Method=Ties, Base Model=Meta-llama-3.1-instruct-8b2025.02 | 24 | 17.59 | |
| SFT on allFinetune Method=SFT on all, Base Model=Meta-llama-3.1-instruct-8b2025.02 | 22.46 | 14.33 | |
| TwinFinetune Method=Twin, Base Model=Meta-llama-3.1-instruct-8b2025.02 | 22.22 | 14.97 | |
| w/o tuningFinetune Method=w/o tuning, Base Model=Meta-llama-3.1-instruct-8b2025.02 | 22.18 | 13.69 | |
| SFT on cluster 0Finetune Method=SFT on cluster 0, Base Model=Meta-llama-3.1-instruct-8b2025.02 | 22.1 | 12.41 | |
| DareFinetune Method=Dare, Base Model=Meta-llama-3.1-instruct-8b2025.02 | 21.92 | 14.28 |