Mathematical Reasoning on GSM-Symbolic (Robustness Analysis)
90.2Accuracy (Vary Both)AbstRaL
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| AbstRaLModel=Qwen2.5-Math-7B-Instruct2025.06 | 90.2 | 0.016 | 91 | 0.86 | |
| CoT-RLModel=Qwen2.5-Math-7B-Instruct2025.06 | 88.1 | 0.021 | 96 | 8.21 | |
| CoT-8SModel=Qwen2.5-Math-7B-Instruct2025.06 | 87.7 | 0.023 | 95 | 7.73 | |
| SyReLMModel=Qwen2.5-Math-7B-Instruct2025.06 | 87.7 | 0.026 | 90 | 2.56 | |
| PoT-8SModel=Qwen2.5-Math-7B-Instruct2025.06 | 86.7 | 0.023 | 91 | 4.77 | |
| AoTModel=Qwen2.5-Math-7B-Instruct2025.06 | 70.4 | 0.032 | 72 | 2.17 | |
| CoAModel=Qwen2.5-Math-7B-Instruct2025.06 | 69.7 | 0.033 | 71 | 1.83 | |
| AbstRaLModel=Qwen2.5-0.5B-Instruct2025.06 | 44.6 | 0.025 | 44 | -1.27 | |
| SyReLMModel=Qwen2.5-0.5B-Instruct2025.06 | 36.8 | 0.03 | 39 | 5.54 | |
| CoT-8SModel=Qwen2.5-0.5B-Instruct2025.06 | 34 | 0.033 | 38 | 10.6 | |
| CoT-RLModel=Qwen2.5-0.5B-Instruct2025.06 | 32.3 | 0.032 | 35 | 7.77 | |
| CoAModel=Qwen2.5-0.5B-Instruct2025.06 | 28.7 | 0.026 | 29 | 0.97 | |
| AoTModel=Qwen2.5-0.5B-Instruct2025.06 | 27.8 | 0.029 | 29 | 4.21 | |
| PoT-8SModel=Qwen2.5-0.5B-Instruct2025.06 | 27 | 0.035 | 31 | 12.8 |