Question Answering on ContextBench Single-Document Ours
91.5AccuracyDeepRead
Evaluation Results
| Method | Links | |
|---|---|---|
| DeepRead2026.02 | 91.5 | |
| DeepReadSearch expansion=Enabled2026.02 | 88.3 | |
| Search-o1Search expansion=Enabled2026.02 | 84 | |
| Search-o12026.02 | 74.5 | |
| ITRGRetrieval refinement strategy=refine2026.02 | 70.1 | |
| ITRGRetrieval refinement strategy=refresh2026.02 | 67.9 | |
| Dense RAGReranking=Enabled2026.02 | 66.2 | |
| Dense RAG2026.02 | 60 | |
| RAPTOR2026.02 | 52.6 |