Information Seeking on DeepWide Search Benchmark
55.9Col-F1Claude-Sonnet-4 (TaS)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Claude-Sonnet-4 (TaS)ReAct Type=MA2026.02 | 55.9 | 63.5 | |
| Claude-Sonnet-4 (TaS) + 32B Sub-AgentReAct Type=MA2026.02 | 52.7 | 67.7 | |
| Gemini DeepResearch2026.02 | 51.2 | 58.3 | |
| Claude-Sonnet-4ReAct Type=SA2026.02 | 39.5 | 35.2 | |
| Claude-Sonnet-4ReAct Type=MA2026.02 | 39.3 | 44.2 |