End-to-end Question Answering on DoRA full (test)
56.53Token F1 ScoreDoRA SFT
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| DoRA SFTRetriever=GTE-multilingual-Base, Retriever top-k=3, Training Protocol=LoRA SFT2026.04 | 56.53 | 56.52 | 40.56 | 0.015 | 89.54 | 66.59 | 3.5 | 29.91 | |
| GPT-3.5-TurboRetriever=GTE-multilingual-Base, Retriever top-k=32026.04 | 38.31 | 39.07 | 14.42 | -0.412 | 83.34 | 56.71 | 7.21 | 36.08 | |
| Llama3-8B-InstructRetriever=GTE-multilingual-Base, Retriever top-k=32026.04 | 38.27 | 38.96 | 14.67 | -0.417 | 83.31 | 55.83 | 7.13 | 37.04 | |
| GPT-4o-miniRetriever=GTE-multilingual-Base, Retriever top-k=32026.04 | 38.24 | 38.92 | 14.61 | -0.413 | 83.34 | 56.5 | 7.34 | 36.15 | |
| Llama3.1-8B-Instruct (base)Retriever=GTE-multilingual-Base, Retriever top-k=32026.04 | 38.13 | 38.88 | 14.46 | -0.416 | 83.3 | 56.23 | 7.04 | 36.73 | |
| Claude-3-HaikuRetriever=GTE-multilingual-Base, Retriever top-k=32026.04 | 38 | 38.7 | 14.77 | -0.429 | 83.19 | 55.89 | 6.63 | 37.48 | |
| Mistral-7B-InstructRetriever=GTE-multilingual-Base, Retriever top-k=32026.04 | 37.98 | 38.65 | 14.59 | -0.42 | 83.2 | 55.83 | 6.65 | 37.52 | |
| GPT-4oRetriever=GTE-multilingual-Base, Retriever top-k=32026.04 | 37.89 | 38.67 | 14.72 | -0.426 | 83.18 | 55.85 | 7.51 | 36.64 | |
| Qwen2.5-7B-InstructRetriever=GTE-multilingual-Base, Retriever top-k=32026.04 | 37.74 | 38.5 | 14.65 | -0.425 | 83.2 | 55.68 | 7.09 | 37.23 |