Single-Hop Question Answering on Natural Questions (NQ)
64Avg@4OWL-8B
Evaluation Results
| Method | Links | |
|---|---|---|
| OWL-8BSetting=Multi-Agent System2026.02 | 64 | |
| SingleSeek-R1-4BSetting=Single Agent2026.02 | 58.8 | |
| AgentFlow-7BSetting=Multi-Agent System2026.02 | 58.5 | |
| WIDESEEK-R1-4BSetting=Multi-Agent System2026.02 | 56.1 | |
| ASearcher-7BSetting=Single Agent2026.02 | 54.5 | |
| MiroFlow-8BSetting=Multi-Agent System2026.02 | 50.9 | |
| Search-R1-7BSetting=Single Agent2026.02 | 49.9 | |
| Qwen3-4BSetting=Multi-Agent System2026.02 | 49.6 | |
| Qwen3-4BSetting=Single Agent2026.02 | 48.5 |