Multi-domain Reasoning on AFlow-Stream (test)
98GSM8K AccuracyEVOCHAMBER
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| EVOCHAMBERBackbone=Qwen3-8B, Team size (k)=32026.05 | 98 | 89.5 | 84.3 | 82 | 83 | 86 | 87.1 | — | |
| AgentNetBackbone=Qwen3-8B, Team size (k)=32026.05 | 97 | 82 | 79.3 | 68 | 90 | 80 | 82.7 | — | |
| SABackbone=Qwen3-8B, Team size (k)=12026.05 | 96 | 79.1 | 78 | 78 | 80 | 80 | 81.9 | — | |
| MemCollabBackbone=Qwen3-8B, Team size (k)=32026.05 | 96 | 84.7 | 79.3 | 66 | 89 | 84 | 83.2 | — | |
| MemCollabBackbone=GPT-4.1-mini2026.05 | 95 | 86.4 | 91 | 68 | 94 | 85 | 86.6 | 0.003 | |
| EVOCHAMBERBackbone=GPT-4.1-mini2026.05 | 95 | 87.8 | 96 | 82 | 94 | 78 | 88.8 | 0.019 | |
| SABackbone=GPT-4.1-mini2026.05 | 94 | 84.7 | 88.7 | 80 | 94 | 80 | 86.9 | — | |
| EvoMemBackbone=GPT-4.1-mini2026.05 | 94 | 89.6 | 91 | 68 | 95 | 86 | 87.3 | 0.004 | |
| EvoMemBackbone=Qwen3-8B, Team size (k)=32026.05 | 94 | 89.2 | 81.7 | 66 | 88 | 85 | 84 | — | |
| SC (k=5)Backbone=Qwen3-8B, Team size (k)=52026.05 | 89 | 77.8 | 56 | 61 | 41 | 69 | 65.6 | — | |
| DyLANBackbone=Qwen3-8B, Team size (k)=32026.05 | 67 | 88.8 | 69 | 62 | 83 | 84 | 75.6 | — |