Section Retrieval on NaturalQuestions (250 questions)
61.4Suppress RateEvolve (Warm)
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Evolve (Warm)Stage=Warm, Model=Qwen 3.5 2B, Threshold=0.802026.04 | 61.4 | 61 | 59.2 | 35.2 | 0.79 | 8.3 | |
| Evolve (Post-consolidation)Stage=Post-consolidation, Model=Qwen 3.5 2B, Threshold=0.802026.04 | 60.8 | 61.4 | 57.6 | 38.8 | 0.74 | 8.1 | |
| Evolve (Cold)Stage=Cold, Model=Qwen 3.5 2B, Threshold=0.802026.04 | 60.2 | 60 | 57.2 | 0 | 1.15 | 9.7 | |
| BaselineStage=Baseline, Model=Qwen 3.5 2B, Threshold=0.802026.04 | 20.4 | — | — | — | 0 | 0.9 |