Retrieval Following on ConFiQA MR 1.0 (test)
61.2PcContext-DPO
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Context-DPOModel=QWEN2-7B-INSTRUCT2024.12 | 61.2 | 20.9 | 24.5 | 27.7 | |
| Context-DPOModel=LLAMA3-8B-INSTRUCT2024.12 | 54.6 | 21.3 | 28.1 | 16 | |
| Context-DPOModel=LLAMA2-7B-CHAT2024.12 | 54.3 | 11.3 | 17.3 | 32.7 | |
| O&IModel=LLAMA2-7B-CHAT2024.12 | 52 | 16 | 23.5 | 29.3 | |
| Context-DPOModel=MISTRAL-7B-INSTRUCT2024.12 | 48.7 | 20.3 | 29.5 | 11.5 | |
| SFTModel=LLAMA2-7B-CHAT2024.12 | 46.7 | 25.5 | 35.3 | 0.4 | |
| BaseModel=LLAMA2-7B-CHAT2024.12 | 45.4 | 26.8 | 37.1 | 0.3 | |
| AttrModel=LLAMA2-7B-CHAT2024.12 | 45.3 | 27.3 | 37.6 | 0.1 | |
| O&IModel=QWEN2-7B-INSTRUCT2024.12 | 43.3 | 43.7 | 43 | 4.7 | |
| O&IModel=MISTRAL-7B-INSTRUCT2024.12 | 33 | 31 | 48.4 | 5.3 | |
| SFTModel=LLAMA3-8B-INSTRUCT2024.12 | 32.7 | 42.5 | 56.5 | 0 | |
| BaseModel=LLAMA3-8B-INSTRUCT2024.12 | 30.6 | 44.1 | 59.1 | 0 | |
| AttrModel=QWEN2-7B-INSTRUCT2024.12 | 26.3 | 40 | 60.3 | 0 | |
| AttrModel=MISTRAL-7B-INSTRUCT2024.12 | 24.9 | 34.7 | 58.2 | 0.7 | |
| O&IModel=LLAMA3-8B-INSTRUCT2024.12 | 24 | 30.7 | 54.1 | 0.7 | |
| SFTModel=MISTRAL-7B-INSTRUCT2024.12 | 22.1 | 37.7 | 63.1 | 0.6 | |
| SFTModel=QWEN2-7B-INSTRUCT2024.12 | 21.9 | 48 | 68.7 | 0 | |
| BaseModel=MISTRAL-7B-INSTRUCT2024.12 | 21.7 | 37.9 | 63.5 | 0.2 | |
| BaseModel=QWEN2-7B-INSTRUCT2024.12 | 21.7 | 48.7 | 69.2 | 0 | |
| AttrModel=LLAMA3-8B-INSTRUCT2024.12 | 20 | 46.3 | 69.9 | 0 |