Open-domain Question Answering on NQ-Open OOD (test)
82.81Exact Match (EM)SkillFlow
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SkillFlowMethod Type=Ours, Backbone Model=Qwen3.5-9B2026.05 | 82.81 | 86.04 | |
| FlowSteerMethod Type=Agent+RL, Backbone Model=Qwen3.5-9B2026.05 | 80.47 | 84.88 | |
| AgentFlowMethod Type=Agent+RL, Backbone Model=Qwen3.5-9B2026.05 | 78.91 | 83.44 | |
| SkillRLMethod Type=Agent+RL, Backbone Model=Qwen3.5-9B2026.05 | 78.91 | 83.09 | |
| AFlowMethod Type=Search-based workflow, Backbone Model=Qwen3.5-9B2026.05 | 75.79 | 80.23 | |
| v4-flashMethod Type=Baseline, Backbone Model=v4-flash2026.05 | 37.5 | 47.9 | |
| Qwen3.5Method Type=SFT, Backbone Model=Qwen3.5-9B2026.05 | 23.44 | 30.97 | |
| Qwen3.5Method Type=Baseline, Backbone Model=Qwen3.5-9B2026.05 | 21.88 | 30.03 | |
| Qwen3.5Method Type=GRPO, Backbone Model=Qwen3.5-9B2026.05 | 21.88 | 28.84 |