Reasoning on MATH, GSM8K, AQUA, GSM-H, MMLU, MMLU-P, GPQA, AIME
84.4MATH AccuracySELFORG
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| SELFORGBackbone=Qwen-2.5-72B-Instruct2025.10 | 84.4 | 96.2 | 80.71 | 64.2 | 83.8 | 71.2 | 47.77 | 23.33 | 68.95 | 1.38 | |
| SingleBackbone=Qwen-2.5-72B-Instruct2025.10 | 83 | 95 | 81.1 | 63.8 | 82.4 | 70.6 | 46.65 | 20 | 67.82 | 2.88 | |
| CoTBackbone=Qwen-2.5-72B-Instruct2025.10 | 82.8 | 95.2 | 80.71 | 62 | 82.8 | 71.4 | 44.2 | 16.67 | 66.97 | 3.5 | |
| AgentVerseBackbone=Qwen-2.5-72B-Instruct2025.10 | 82.8 | 95.2 | 77.17 | 57.8 | 81.4 | 71.2 | 45.98 | 23.33 | 66.86 | 4.13 | |
| MacNetBackbone=Qwen-2.5-72B-Instruct2025.10 | 81.4 | 95.4 | 79.13 | 62.8 | 83.2 | 65.6 | 40.4 | 16.67 | 65.58 | 4.13 | |
| AutoGenBackbone=Qwen-2.5-72B-Instruct2025.10 | 81.2 | 95.8 | 78.35 | 64.2 | 82.6 | 69.4 | 45.54 | 13.33 | 66.3 | 3.75 | |
| DyLANBackbone=Qwen-2.5-72B-Instruct2025.10 | 80.6 | 95.4 | 77.95 | 63.2 | 84.2 | 69.2 | 46.43 | 13.33 | 66.29 | 3.75 | |
| SELFORGBackbone=LLaMA-3.3-70B-Instruct2025.10 | 79.8 | 96.6 | 81.1 | 56.8 | 85 | 72.4 | 59.82 | 30 | 70.19 | 1.25 | |
| DyLANBackbone=LLaMA-3.3-70B-Instruct2025.10 | 77.6 | 95.2 | 76.38 | 53 | 83.6 | 31.6 | 58.04 | 26.67 | 62.76 | 4.25 | |
| AgentVerseBackbone=LLaMA-3.3-70B-Instruct2025.10 | 76.8 | 94.6 | 76.38 | 51.2 | 83.6 | 69.2 | 55.36 | 26.67 | 66.73 | 4.5 | |
| CoTBackbone=LLaMA-3.3-70B-Instruct2025.10 | 75 | 95.8 | 79.92 | 57.4 | 85.2 | 71 | 56.7 | 26.67 | 68.46 | 2.5 | |
| SingleBackbone=LLaMA-3.3-70B-Instruct2025.10 | 74.8 | 96.2 | 77.56 | 54 | 84.4 | 68.4 | 55.36 | 23.33 | 66.76 | 3.88 | |
| MacNetBackbone=LLaMA-3.3-70B-Instruct2025.10 | 74.8 | 96 | 79.13 | 55.2 | 83 | 65.4 | 58.26 | 26.67 | 67.31 | 3.63 | |
| AutoGenBackbone=LLaMA-3.3-70B-Instruct2025.10 | 70.8 | 93 | 79.5 | 51.4 | 82.6 | 64.6 | 52.68 | 30 | 65.57 | 5.13 |