Question Answering on HotpotQA fixed sample of 200 (dev)
26.62F1 Scorereact
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| react2026.03 | 26.62 | 546.6 | 0.56 | 0.0005 | |
| policy (step-cost)cost_proxy=step_cost2026.03 | 23.6 | 1,294.2 | 1.138 | 0.0002 | |
| workflow-search-twice2026.03 | 16.98 | 514.1 | 0.902 | 0.0003 | |
| workflow (minimal)2026.03 | 16.25 | 451.2 | 0.461 | 0.0004 | |
| threshold2026.03 | 12.55 | 350.3 | 0.394 | 0.0004 | |
| direct2026.03 | 7.19 | 93 | 0.122 | 0.0008 | |
| workflow-search-verify2026.03 | 6.3 | 1,041.2 | 1.617 | 0.0001 |