Long document retrieval on ConditionalQA (test)
35.26F1 ScoreAttentionRetriever
Evaluation results
| Method | Links | |
|---|---|---|
| AttentionRetrieverBackbone=LLaMA-3.2 3B2026.02 | 35.26 | |
| GritLMModel Type=Dense2026.02 | 32.58 | |
| AttentionRetrieverBackbone=Qwen-2.5 3B2026.02 | 31.06 | |
| Qwen3Model Type=Dense2026.02 | 28.92 | |
| GTRModel Type=Dense2026.02 | 25.84 | |
| GTE-Qwen2Model Type=Dense2026.02 | 24.15 | |
| BM25Model Type=Sparse2026.02 | 19.88 | |
| ANCEModel Type=Dense2026.02 | 18.93 | |
| CDEModel Type=Dense2026.02 | 15.47 | |
| DPRModel Type=Dense2026.02 | 15.42 | |
| SPScannerModel Type=Autoregressive2026.02 | 13.54 |