Multi-tool calling on VitaBench (held-out test)
63.1Success RateSkill-MAS-optimized
Evaluation Results
| Method | Links | |
|---|---|---|
| Skill-MAS-optimizedMeta-agent=DeepSeek-V4-Flash2026.06 | 63.1 | |
| EvoAgentMeta-agent=DeepSeek-V4-Flash2026.06 | 61.9 | |
| AOrchestraMeta-agent=DeepSeek-V4-Flash2026.06 | 58.33 | |
| AFlowMeta-agent=DeepSeek-V4-Flash2026.06 | 55.95 | |
| Skill-MAS-initMeta-agent=DeepSeek-V4-Flash2026.06 | 53.57 | |
| Skill-MAS-optimizedMeta-agent=Qwen3.5-Plus2026.06 | 48.8 | |
| MAS-OrchestraMeta-agent=DeepSeek-V4-Flash2026.06 | 47.62 | |
| AFlowMeta-agent=Qwen3.5-Plus2026.06 | 45.24 | |
| AOrchestraMeta-agent=Qwen3.5-Plus2026.06 | 40.48 | |
| Skill-MAS-initMeta-agent=Qwen3.5-Plus2026.06 | 36.9 | |
| MAS-OrchestraMeta-agent=Qwen3.5-Plus2026.06 | 35.71 | |
| EvoAgentMeta-agent=Qwen3.5-Plus2026.06 | 34.52 | |
| MAS2Meta-agent=DeepSeek-V4-Flash2026.06 | 33.33 | |
| Skill-MAS-optimizedMeta-agent=Gemini-3.1-Flash2026.06 | 28.6 | |
| MAS2Meta-agent=Qwen3.5-Plus2026.06 | 22.62 | |
| AFlowMeta-agent=Gemini-3.1-Flash2026.06 | 19.05 | |
| Skill-MAS-initMeta-agent=Gemini-3.1-Flash2026.06 | 16.67 | |
| Skill-MAS-optimizedMeta-agent=GPT-5.4-Nano2026.06 | 15.48 | |
| AOrchestraMeta-agent=Gemini-3.1-Flash2026.06 | 14.29 | |
| EvoAgentMeta-agent=Gemini-3.1-Flash2026.06 | 10.71 | |
| MAS-OrchestraMeta-agent=Gemini-3.1-Flash2026.06 | 9.52 | |
| MAS-OrchestraMeta-agent=GPT-5.4-Nano2026.06 | 9.52 | |
| MAS2Meta-agent=Gemini-3.1-Flash2026.06 | 8.33 | |
| EvoAgentMeta-agent=GPT-5.4-Nano2026.06 | 8.33 | |
| Skill-MAS-initMeta-agent=GPT-5.4-Nano2026.06 | 5.95 | |
| AOrchestraMeta-agent=GPT-5.4-Nano2026.06 | 4.76 | |
| AFlowMeta-agent=GPT-5.4-Nano2026.06 | 2.38 | |
| MAS2Meta-agent=GPT-5.4-Nano2026.06 | 2.38 |