Multi-document Retrieval on HotpotQA, 2WikiMultihopQA, and MuSiQue (val)
64.61Average F1GritLM
Evaluation Results
| Method | Links | |
|---|---|---|
| GritLMModel Category=Dense Models, top_k=32026.02 | 64.61 | |
| GTE-Qwen2Model Category=Dense Models, top_k=32026.02 | 62.25 | |
| AttentionRetrieverModel Category=AttentionRetriever (Ours), Backbone=LLaMA-3.2, Model size=3B, top_k=32026.02 | 62.23 | |
| Qwen3Model Category=Dense Models, top_k=32026.02 | 61.58 | |
| AttentionRetrieverModel Category=AttentionRetriever (Ours), Backbone=Qwen-2.5, Model size=3B, top_k=32026.02 | 61.51 | |
| GTRModel Category=Dense Models, top_k=32026.02 | 56.98 | |
| ANCEModel Category=Dense Models, top_k=32026.02 | 54.78 | |
| SPScannerModel Category=Autoregressive Models, top_k=32026.02 | 54.34 | |
| BM25Model Category=Sparse Models, top_k=32026.02 | 48.54 | |
| DPRModel Category=Dense Models, top_k=32026.02 | 46.96 | |
| CDEModel Category=Dense Models, top_k=32026.02 | 29.63 |