notesum.ai
Published at October 22Self-Evolving Multi-Agent Collaboration Networks for Software Development
cs.AI
cs.GT
cs.LG
Released Date: October 22, 2024
Authors: Yue Hu1, Yuzhu Cai2, Yaxin Du1, Xinyu Zhu1, Xiangrui Liu1, Zijie Yu1, Yuchen Hou1, Shuo Tang1, Siheng Chen3
Aff.: 1Shanghai Jiao Tong University; 2Beihang University; 3Shanghai Jiao Tong University, Shanghai AI Laboratory

| rSDE-Bench | HumanEval (%) | |||||
| Website(%) | Game(%) | |||||
| Method | Model | Basic | Advanced | Basic | Advanced | Pass@1 |
| Gemini-1.5-Flash | 29.79±1.00 | 11.61±2.34 | 21.74±6.39 | 6.45±6.97 | 73.17 | |
| Claude-3.5-Sonnet | 58.90±1.48 | 37.11±1.06 | 44.20±5.41 | 18.29±13.26 | 89.02 | |
| Single-Agent | GPT-4o-Mini | 62.90±2.52 | 44.40±4.21 | 42.76±15.50 | 30.10±11.87 | 88.41 |
| MetaGPT | 15.41±0.00 | 0.00±0.00 | 16.67±2.71 | 0.00±0.00 | 88.41 | |
| Autogen | 25.68±4.14 | 5.40±3.34 | 17.39±1.78 | 0.00±0.00 | 85.36 | |
| MapCoder | 34.70±1.59 | 14.57±0.66 | 29.71±6.72 | 7.52±6.10 | 90.85 | |
| Agentverse | 15.41±0.00 | 0.00±0.00 | 37.67±8.20 | 16.13±4.55 | 90.85 | |
| Multi-Agent | ChatDev | 62.67±0.28 | 43.45±0.77 | 53.63±5.70 | 32.26±4.55 | 70.73 |
| 89.38±1.01 | 65.05±1.56 | 77.54±2.04 | 51.60±4.54 | 94.51 | ||
| EvoMAC | +26.48 | +20.65 | +34.78 | +21.50 | +6.10 | |