Search-Augmented Reasoning on SEARCHGEN-BENCH NoSearch
56.9ScoreGENERATOR-ADAPTIVE SEARCH (RFT-8B) + Klein-4B-DPO
Evaluation Results
| Method | Links | |
|---|---|---|
| GENERATOR-ADAPTIVE SEARCH (RFT-8B) + Klein-4B-DPOPhase=2, Architecture=Klein-4B, Search Strategy=GENERATOR-ADAPTIVE SEARCH, Reasoner Config=RFT-8B2026.07 | 56.9 | |
| ORACLE + Klein-4B-DPOArchitecture=Klein-4B, Search Strategy=ORACLE2026.07 | 55.7 | |
| BLIND SEARCH (SFT-8B) + Klein-4BPhase=0, Architecture=Klein-4B, Search Strategy=BLIND SEARCH, Reasoner Config=SFT-8B2026.07 | 54.6 | |
| GENERATOR-ADAPTIVE SEARCH (RFT-8B) + Klein-4BArchitecture=Klein-4B, Search Strategy=GENERATOR-ADAPTIVE SEARCH, Reasoner Config=RFT-8B2026.07 | 54.6 | |
| GENERATOR-ADAPTIVE SEARCH (RFT-8B) + Bagel-DPOPhase=2, Architecture=Bagel, Search Strategy=GENERATOR-ADAPTIVE SEARCH, Reasoner Config=RFT-8B2026.07 | 54.3 | |
| BLIND SEARCH (SFT-8B) + Klein-4B-DPOPhase=1, Architecture=Klein-4B, Search Strategy=BLIND SEARCH, Reasoner Config=SFT-8B2026.07 | 54 | |
| ORACLE + Bagel-DPOArchitecture=Bagel, Search Strategy=ORACLE2026.07 | 53.2 | |
| BLIND SEARCH (SFT-8B) + BagelPhase=0, Architecture=Bagel, Search Strategy=BLIND SEARCH, Reasoner Config=SFT-8B2026.07 | 52.6 | |
| GENERATOR-ADAPTIVE SEARCH (RFT-8B) + BagelArchitecture=Bagel, Search Strategy=GENERATOR-ADAPTIVE SEARCH, Reasoner Config=RFT-8B2026.07 | 52.5 | |
| BLIND SEARCH (SFT-8B) + Bagel-DPOPhase=1, Architecture=Bagel, Search Strategy=BLIND SEARCH, Reasoner Config=SFT-8B2026.07 | 52.4 | |
| NO SEARCH + Klein-4B-DPOArchitecture=Klein-4B, Search Strategy=NO SEARCH2026.07 | 49.9 | |
| NO SEARCH + Bagel-DPOArchitecture=Bagel, Search Strategy=NO SEARCH2026.07 | 48.8 |