Open-domain Question Answering on TriviaQA, NQ Open, WebQS Foundation-Preservation Suite
52.51TriviaQA ScoreLLaMA-2-7B
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| LLaMA-2-7B#Param=-, Base Model=LLaMA-2-7B, Fine-tuning Dataset=CodeFeedback2026.05 | 52.51 | 16.91 | 5.88 | 25.1 | 100 | — | |
| LoRA#Param=320M, Base Model=LLaMA-2-7B, Fine-tuning Dataset=CodeFeedback2026.05 | 51.36 | 10.77 | 8.56 | 23.57 | 93.87 | 21.26 | |
| FoLoRA#Param=320M, Base Model=LLaMA-2-7B, Fine-tuning Dataset=CodeFeedback2026.05 | 51.31 | 14.78 | 10.82 | 25.63 | 102.12 | 23.87 | |
| CorDA#Param=320M, Base Model=LLaMA-2-7B, Fine-tuning Dataset=CodeFeedback2026.05 | 49.32 | 12.72 | 6.72 | 22.92 | 91.33 | 20.85 | |
| MiLoRA#Param=320M, Base Model=LLaMA-2-7B, Fine-tuning Dataset=CodeFeedback2026.05 | 49.28 | 11.8 | 7.1 | 22.72 | 90.54 | 20.7 | |
| LoRA-Null#Param=320M, Base Model=LLaMA-2-7B, Fine-tuning Dataset=CodeFeedback2026.05 | 49.09 | 13.78 | 6.61 | 23.16 | 92.27 | 21.97 | |
| OPLoRA#Param=320M, Base Model=LLaMA-2-7B, Fine-tuning Dataset=CodeFeedback2026.05 | 44.32 | 9.07 | 6.1 | 19.83 | 79 | 20.02 |