Multi-hop dynamic question answering on BrowseComp-Plus (test)
27.38Accuracy (%)Skill-MAS-optimized
Evaluation Results
| Method | Links | |
|---|---|---|
| Skill-MAS-optimizedMeta-agent=GPT-5.4-Nano2026.06 | 27.38 | |
| EvoAgentMeta-agent=GPT-5.4-Nano2026.06 | 23.81 | |
| Skill-MAS-optimizedMeta-agent=Qwen3.5-Plus2026.06 | 23.81 | |
| Skill-MAS-optimizedMeta-agent=Gemini-3.1-Flash2026.06 | 23.21 | |
| AOrchestraMeta-agent=GPT-5.4-Nano2026.06 | 22.62 | |
| Skill-MAS-optimizedMeta-agent=DeepSeek-V4-Flash2026.06 | 22.62 | |
| Skill-MAS-initMeta-agent=Qwen3.5-Plus2026.06 | 20.24 | |
| Skill-MAS-initMeta-agent=GPT-5.4-Nano2026.06 | 19.64 | |
| Skill-MAS-initMeta-agent=Gemini-3.1-Flash2026.06 | 19.05 | |
| AOrchestraMeta-agent=Qwen3.5-Plus2026.06 | 19.05 | |
| AFlowMeta-agent=GPT-5.4-Nano2026.06 | 18.45 | |
| AOrchestraMeta-agent=DeepSeek-V4-Flash2026.06 | 18.45 | |
| AFlowMeta-agent=Gemini-3.1-Flash2026.06 | 17.86 | |
| EvoAgentMeta-agent=DeepSeek-V4-Flash2026.06 | 16.67 | |
| EvoAgentMeta-agent=Gemini-3.1-Flash2026.06 | 16.07 | |
| MAS-OrchestraMeta-agent=DeepSeek-V4-Flash2026.06 | 15.72 | |
| EvoAgentMeta-agent=Qwen3.5-Plus2026.06 | 15.48 | |
| AFlowMeta-agent=DeepSeek-V4-Flash2026.06 | 15.48 | |
| Skill-MAS-initMeta-agent=DeepSeek-V4-Flash2026.06 | 15.48 | |
| AOrchestraMeta-agent=Gemini-3.1-Flash2026.06 | 14.88 | |
| AFlowMeta-agent=Qwen3.5-Plus2026.06 | 12.5 | |
| MAS2Meta-agent=Qwen3.5-Plus2026.06 | 11.52 | |
| MAS2Meta-agent=Gemini-3.1-Flash2026.06 | 11.39 | |
| MAS2Meta-agent=GPT-5.4-Nano2026.06 | 10.71 | |
| MAS-OrchestraMeta-agent=Qwen3.5-Plus2026.06 | 10.12 | |
| MAS-OrchestraMeta-agent=Gemini-3.1-Flash2026.06 | 8.9 | |
| MAS-OrchestraMeta-agent=GPT-5.4-Nano2026.06 | 7.74 | |
| MAS2Meta-agent=DeepSeek-V4-Flash2026.06 | 4.92 |