Literature Search on ROLLINGEVAL JUN25
67.9Semantic Relevance@10Debate+QWEN3
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Debate+QWEN3Approach=Ensemble, Components=Debate + QWEN32026.05 | 67.9 | 53.2 | 76.2 | 87.9 | 53.8 | 33.2 | |
| QWEN3 emb.Approach=Embedding similarity2026.05 | 67.1 | 47.6 | 65.3 | 77.8 | 57.6 | 47.9 | |
| DebateApproach=Prompt-based scoring2026.05 | 55.4 | 54.1 | 79.7 | 89.1 | 39.8 | 27.1 | |
| Human ReferencesType=Ground-truth benchmark2026.05 | 47.3 | — | — | — | 50.7 | 50.8 | |
| No Rerank (no DR)Deep Research Pipeline=False2026.05 | 9.2 | 15.3 | 41.2 | 62.1 | 17.3 | 17.4 |