Code Generation on MBPP+, CC, HumanEval
88.7MBPP+ AccuracyAgentNet
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| AgentNetBackbone=Qwen3-8B, Team size (k)=32026.05 | 88.7 | 10.2 | 69.8 | |
| EvoMemBackbone=Qwen3-8B, Team size (k)=32026.05 | 88.5 | 2.7 | 67.2 | |
| MemCollabBackbone=Qwen3-8B, Team size (k)=32026.05 | 87 | 8.4 | 68.2 | |
| EVOCHAMBERBackbone=Qwen3-8B, Team size (k)=3, N=202026.05 | 86.1 | 35.2 | 75.7 | |
| SC (k=5)Backbone=Qwen3-8B, Team size (k)=52026.05 | 84.9 | 19.8 | 70.8 | |
| SABackbone=Qwen3-8B, Team size (k)=12026.05 | 84.2 | 6.8 | 66.7 | |
| DyLANBackbone=Qwen3-8B, Team size (k)=32026.05 | 82.5 | 18.9 | 69.5 |