Open-Domain Question Answering on Bamboogle
18.4Joint Answer-Evidence ScoreInitial (no search)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Initial (no search)search=false2026.05 | 18.4 | — | |
| EVE-Agentsearch=true2026.05 | 6.4 | — | |
| Dr. Zerosearch=true2026.05 | 4 | — | |
| Initial (search)search=true2026.05 | 2.4 | — | |
| HiPRAGBackbone=Qwen2.5-7B-Instruct2026.05 | 2.23 | 47.2 | |
| HiPRAGBackbone=Qwen2.5-3B-Instruct2026.05 | 2.08 | 36.8 | |
| StepSearchBackbone=Qwen2.5-7B-Instruct2026.05 | 1.84 | 49.6 | |
| StepSearchBackbone=Qwen2.5-3B-Instruct2026.05 | 1.7 | 27.2 | |
| Search-R1Backbone=Qwen2.5-3B-Instruct2026.05 | 1.53 | 40.8 | |
| RFTBackbone=Qwen2.5-3B-Instruct2026.05 | 1.4 | 40.8 | |
| RFTBackbone=Qwen2.5-7B-Instruct2026.05 | 1.36 | 42.4 | |
| SAASBackbone=Qwen2.5-3B-Instruct2026.05 | 1.23 | 44.8 | |
| Search-R1Backbone=Qwen2.5-7B-Instruct2026.05 | 1.15 | 40 | |
| SAASBackbone=Qwen2.5-7B-Instruct2026.05 | 1.02 | 49.6 | |
| Direct InferenceBackbone=Qwen2.5-3B-Instruct2026.05 | — | 28.8 | |
| Direct InferenceBackbone=Qwen2.5-7B-Instruct2026.05 | — | 40 |