Multi-hop Retrieval on MultiHop-RAG (test)
63.26NDCG@10ScalDPP
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| ScalDPPBackbone=Qwen/Qwen3-Embedding-4B, Reranker=BAAI/bge-reranker-v2-m32026.02 | 63.26 | 78.55 | 81.23 | 59.8 | 69.07 | 74.97 | |
| ScalDPPBackbone=BAAI/bge-m3, Reranker=BAAI/bge-reranker-v2-m32026.02 | 60.89 | 74.8 | 77.65 | 57.19 | 65.02 | 72.4 | |
| ScalDPPBackbone=BAAI/bge-large-en-v1.5, Reranker=BAAI/bge-reranker-v2-m32026.02 | 60.7 | 72.1 | 74.53 | 58.15 | 65.25 | 69.83 | |
| Standard RAGBackbone=Qwen/Qwen3-Embedding-4B, Reranker=BAAI/bge-reranker-v2-m32026.02 | 60.36 | 77.32 | 80.45 | 54.9 | 62.85 | 70.06 | |
| Standard RAGBackbone=BAAI/bge-large-en-v1.5, Reranker=BAAI/bge-reranker-v2-m32026.02 | 59.09 | 73.56 | 75.87 | 54.79 | 62.32 | 67.37 | |
| Standard RAGBackbone=BAAI/bge-m3, Reranker=BAAI/bge-reranker-v2-m32026.02 | 59.05 | 73.44 | 76.98 | 54.54 | 61.59 | 68.6 | |
| ScalDPPBackbone=Qwen/Qwen3-Embedding-0.6B, Reranker=BAAI/bge-reranker-v2-m32026.02 | 58.84 | 71.73 | 74.19 | 55.98 | 64.13 | 69.83 | |
| Standard RAGBackbone=Qwen/Qwen3-Embedding-0.6B, Reranker=BAAI/bge-reranker-v2-m32026.02 | 57.86 | 73.39 | 76.09 | 53.44 | 61.57 | 66.82 | |
| DPP Base, no adapterBackbone=BAAI/bge-large-en-v1.5, Reranker=BAAI/bge-reranker-v2-m32026.02 | 51.56 | 61.96 | 66.7 | 47.35 | 50.47 | 57.65 | |
| DPP Base, no adapterBackbone=BAAI/bge-m3, Reranker=BAAI/bge-reranker-v2-m32026.02 | 51.54 | 61.53 | 69.39 | 46.89 | 49.39 | 59.78 | |
| DPP Base, no adapterBackbone=Qwen/Qwen3-Embedding-4B, Reranker=BAAI/bge-reranker-v2-m32026.02 | 49.63 | 58.81 | 66.26 | 45.7 | 48.25 | 57.88 | |
| DPP Base, no adapterBackbone=Qwen/Qwen3-Embedding-0.6B, Reranker=BAAI/bge-reranker-v2-m32026.02 | 48.96 | 58.31 | 64.8 | 44.98 | 47.51 | 55.98 | |
| ScalDPPBackbone=Qwen/Qwen3-Embedding-4B, Reranker=None2026.02 | 48.95 | 74.53 | 78.66 | 43.39 | 59.42 | 67.93 | |
| ScalDPPBackbone=BAAI/bge-m3, Reranker=None2026.02 | 46.31 | 71.82 | 76.2 | 39.63 | 53.87 | 63.58 | |
| Standard RAGBackbone=Qwen/Qwen3-Embedding-4B, Reranker=None2026.02 | 45.88 | 66.69 | 72.74 | 37.99 | 45.34 | 54.41 | |
| ScalDPPBackbone=Qwen/Qwen3-Embedding-0.6B, Reranker=None2026.02 | 44.67 | 67.84 | 71.84 | 40.04 | 55.56 | 63.13 | |
| ScalDPPBackbone=BAAI/bge-large-en-v1.5, Reranker=None2026.02 | 43.59 | 69.17 | 72.74 | 38.16 | 54.16 | 61.23 | |
| Standard RAGBackbone=BAAI/bge-m3, Reranker=None2026.02 | 42.59 | 60.5 | 68.27 | 35.45 | 41.42 | 51.62 | |
| Standard RAGBackbone=Qwen/Qwen3-Embedding-0.6B, Reranker=None2026.02 | 41.47 | 61.77 | 67.49 | 34.19 | 41.86 | 50.17 | |
| Standard RAGBackbone=BAAI/bge-large-en-v1.5, Reranker=None2026.02 | 40.53 | 59.23 | 64.47 | 33.54 | 40.46 | 48.6 | |
| DPP Base, no adapterBackbone=Qwen/Qwen3-Embedding-0.6B, Reranker=None2026.02 | 23.29 | 40.56 | 49.94 | 16.45 | 21.84 | 30.17 | |
| DPP Base, no adapterBackbone=BAAI/bge-m3, Reranker=None2026.02 | 22.9 | 42.15 | 53.07 | 14.39 | 19.44 | 27.82 | |
| DPP Base, no adapterBackbone=Qwen/Qwen3-Embedding-4B, Reranker=None2026.02 | 22.65 | 40.52 | 51.17 | 14.97 | 20.47 | 29.72 | |
| DPP Base, no adapterBackbone=BAAI/bge-large-en-v1.5, Reranker=None2026.02 | 21.33 | 41.71 | 48.6 | 13.03 | 18.7 | 24.92 |