Question Answering on Average 7 QA benchmarks
40.54EMRAG-GD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| RAG-GDBackbone=Llama-3.1-8B, Retriever=E5, K=52026.05 | 40.54 | 49.24 | |
| Base adapterBackbone=Llama-3.1-8B, Retriever=E52026.05 | 38.86 | 48.04 | |
| RAG-GDBackbone=Llama-3.1-8B, Retriever=BM25, K=52026.05 | 36.83 | 45.7 | |
| RAG-GDBackbone=Qwen-2.5-7B, Retriever=E5, K=52026.05 | 36.71 | 45.11 | |
| Base adapterBackbone=Llama-3.1-8B, Retriever=BM252026.05 | 35.86 | 45.34 | |
| Base adapterBackbone=Qwen-2.5-7B, Retriever=E52026.05 | 34.16 | 42.54 | |
| Vanilla RAGBackbone=Llama-3.1-8B, Retriever=E52026.05 | 34.01 | 41.9 | |
| Vanilla RAGBackbone=Qwen-2.5-7B, Retriever=E52026.05 | 32.61 | 40.6 | |
| RAG-GDBackbone=Qwen-2.5-7B, Retriever=BM25, K=52026.05 | 32.43 | 40.87 | |
| Base adapterBackbone=Qwen-2.5-7B, Retriever=BM252026.05 | 29.71 | 37.72 | |
| Vanilla RAGBackbone=Llama-3.1-8B, Retriever=BM252026.05 | 28.72 | 36.34 | |
| Vanilla RAGBackbone=Qwen-2.5-7B, Retriever=BM252026.05 | 27.16 | 34.9 | |
| Query OnlyBackbone=Llama-3.1-8B, Retriever=None2026.05 | 21.52 | 28.08 | |
| Query OnlyBackbone=Qwen-2.5-7B, Retriever=None2026.05 | 18.94 | 25.09 |