Multi-agent System Robustness Evaluation on GQA 1.0 (test)
57.7VIAReAct
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| ReActLLM=Qwen-7B2026.05 | 57.7 | 55.2 | 60.8 | 60.7 | 65 | 55.2 | 62.2 | 76.7 | 65.5 | 78.3 | |
| ReActLLM=GLM-4V+2026.05 | 53.5 | 48.3 | 53.7 | 50.3 | 62.2 | 48.3 | 57.5 | 72 | 49.8 | 71.3 | |
| ReActLLM=Qwen-32B2026.05 | 52.5 | 50 | 55.7 | 55.5 | 59.8 | 50 | 57 | 71.5 | 60.3 | 73.2 | |
| ReflexionLLM=Qwen-7B2026.05 | 47.2 | 47.8 | 51.2 | 50.8 | 56.7 | 51.3 | 50.3 | 62.2 | 57.3 | 61.7 | |
| PlanAndSolveLLM=Qwen-7B2026.05 | 46.8 | 53.3 | 59.8 | 53.3 | 71.8 | 62.5 | 56.2 | 69.5 | 60.2 | 69.2 | |
| ReActLLM=O1-Mini2026.05 | 46.3 | 41.2 | 44 | 43.3 | 51.3 | 41.2 | 47.2 | 69.7 | 42.5 | 71.5 | |
| ReflexionLLM=GPT-4o2026.05 | 43 | 42.7 | 34.5 | 43 | 36.2 | 39 | 37.8 | 49.2 | 36.5 | 52.2 | |
| ReflexionLLM=Qwen-32B2026.05 | 42 | 42.7 | 46 | 45.7 | 51.5 | 46.2 | 45.2 | 57 | 52.2 | 56.5 | |
| ReActLLM=GPT-4o2026.05 | 41.8 | 38.2 | 43.2 | 41.8 | 49 | 42.5 | 44.7 | 64.2 | 39.7 | 65 | |
| PlanAndSolveLLM=Qwen-32B2026.05 | 41.7 | 48.2 | 54.7 | 48.2 | 66.7 | 57.3 | 51 | 64.3 | 55 | 64 | |
| ReflexionLLM=GLM-4V+2026.05 | 37.7 | 38.2 | 45 | 39.8 | 46.5 | 42.8 | 41.5 | 53 | 45.7 | 53.3 | |
| PlanAndSolveLLM=GLM-4V+2026.05 | 37.3 | 44.3 | 48 | 44 | 47.3 | 51 | 47.8 | 58.5 | 48.5 | 61 | |
| ReflexionLLM=O1-Mini2026.05 | 33.7 | 33.3 | 37.8 | 33.2 | 38.5 | 36 | 35.5 | 43.5 | 39.3 | 45.2 | |
| PlanAndSolveLLM=O1-Mini2026.05 | 31.7 | 38.3 | 41 | 36.3 | 39.5 | 43.8 | 39.3 | 50.7 | 42.5 | 51.7 | |
| PlanAndSolveLLM=GPT-4o2026.05 | 29.5 | 35.3 | 38.7 | 35.5 | 41.5 | 41.8 | 38.3 | 46.2 | 37.8 | 48.7 |