Retrieval-Augmented Generation on RAG-Bench
87.15F1 (Golden Only)RAAT
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| RAATBackbone=LLaMA2-7B2024.05 | 87.15 | 83.07 | 86.8 | 82.73 | 85.14 | 81 | 86.29 | 82.1 | 86.35 | 82.23 | |
| RALM_multipleBackbone=LLaMA2-7B, Fine-tuning Strategy=Instruction tuning with combined retrieval noise samples2024.05 | 85.47 | 80.17 | 85.27 | 81.2 | 83.07 | 78.33 | 83.25 | 79.23 | 84.27 | 79.73 | |
| RALM_goldenBackbone=LLaMA2-7B, Fine-tuning Strategy=Instruction tuning with golden retrieval text2024.05 | 80.31 | 74.03 | 79.33 | 72.73 | 73.26 | 66.33 | 73.08 | 65.4 | 76.5 | 69.62 | |
| RetRobustBackbone=LLaMA2-7B, Fine-tuning Strategy=Instruction tuning with mixed retrieval noise2024.05 | 80.1 | 73.8 | 79.25 | 72.97 | 74.81 | 68.3 | 75.46 | 68.43 | 77.41 | 70.88 | |
| RALM_retrievedBackbone=LLaMA2-7B, Fine-tuning Strategy=Instruction tuning with top-2 retrieved passages2024.05 | 80.04 | 73.4 | 81.09 | 74.8 | 75.99 | 69.1 | 73.1 | 65.67 | 77.55 | 70.74 | |
| ChatGPT3.52024.05 | 73.98 | 60.5 | 72.24 | 60.3 | 70.65 | 56.89 | 69 | 54.64 | 71.47 | 58.1 | |
| LLaMA2_70BBackbone=LLaMA2, Parameter Size=70B2024.05 | 71.43 | 56.56 | 70.05 | 55.13 | 65.97 | 51.33 | 63.91 | 48.27 | 67.84 | 52.82 | |
| LLaMA2_13BBackbone=LLaMA2, Parameter Size=13B2024.05 | 69.27 | 55 | 63.25 | 49.47 | 62.27 | 47.97 | 62.07 | 47.17 | 64.22 | 49.9 | |
| Qwen_14BBackbone=Qwen, Parameter Size=14B2024.05 | 67.45 | 51.43 | 66.71 | 51.2 | 61.88 | 46.16 | 58.65 | 41.3 | 63.67 | 47.52 | |
| LLaMA2_7BBackbone=LLaMA2, Parameter Size=7B2024.05 | 65.56 | 51.8 | 56.14 | 42.87 | 53.1 | 39.73 | 51.81 | 38.37 | 56.68 | 43.19 | |
| Qwen_7BBackbone=Qwen, Parameter Size=7B2024.05 | 62.57 | 47.07 | 61.48 | 46.06 | 55.5 | 40.5 | 53.26 | 36.9 | 58.2 | 42.63 |