Agentic Search on Agentic-Wide TaskCraft (test)
90.1AccuracyA-MAPREDUCE
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| A-MAPREDUCEBackbone=GPT-5-mini2026.02 | 90.1 | 384.07 | 0.09 | |
| Flash-SearcherBackbone=GPT-5-mini2026.02 | 87.13 | 275.9 | 0.06 | |
| SmolAgentsBackbone=GPT-5-mini2026.02 | 80.2 | 607.04 | 0.12 | |
| OWLBackbone=GPT-5-mini2026.02 | 74.26 | 702.48 | 0.08 |