Mathematical Reasoning on Olympiad (P@1, ASRt, RAS)
40.86P@1BaseRL
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| BaseRLBackbone Model=DeepSeek-R1-Distill-Qwen-1.5B2025.09 | 40.86 | — | — | |
| DecepChainBackbone Model=DeepSeek-R1-Distill-Qwen-1.5B2025.09 | 40.5 | 96.83 | 92.17 | |
| DecepChainBackbone Model=Qwen2.5-Math-7B2025.09 | 36.52 | 98.78 | 96.67 | |
| BadNetBackbone Model=DeepSeek-R1-Distill-Qwen-1.5B2025.09 | 35.17 | 65.16 | 0.93 | |
| BadChainBackbone Model=DeepSeek-R1-Distill-Qwen-1.5B2025.09 | 34.64 | 65.1 | 0 | |
| DTCoTBackbone Model=DeepSeek-R1-Distill-Qwen-1.5B2025.09 | 34.16 | 65.33 | 0 | |
| BaseRLBackbone Model=Qwen2.5-Math-1.5B2025.09 | 32.75 | — | — | |
| BaseRLBackbone Model=Qwen2.5-Math-7B2025.09 | 32.75 | — | — | |
| BadChainBackbone Model=Qwen2.5-Math-7B2025.09 | 31.85 | 68.71 | 1.77 | |
| DecepChainBackbone Model=Qwen2.5-Math-1.5B2025.09 | 31.82 | 96.02 | 87.49 | |
| DTCoTBackbone Model=Qwen2.5-Math-7B2025.09 | 31.76 | 67.32 | 0 | |
| BadNetBackbone Model=Qwen2.5-Math-7B2025.09 | 29.69 | 70.61 | 1 | |
| BadNetBackbone Model=Qwen2.5-Math-1.5B2025.09 | 26.58 | 73.54 | 0.45 | |
| BadNetBackbone Model=Llama-3.2-3B-Instruct2025.09 | 13.48 | 87.76 | 9.23 | |
| BaseRLBackbone Model=Llama-3.2-3B-Instruct2025.09 | 12.59 | — | — | |
| BadChainBackbone Model=Qwen2.5-Math-1.5B2025.09 | 12.12 | 90.61 | 22.49 | |
| DecepChainBackbone Model=Llama-3.2-3B-Instruct2025.09 | 10.91 | 98.19 | 83.38 | |
| DTCoTBackbone Model=Qwen2.5-Math-1.5B2025.09 | 10.79 | 92.56 | 31.04 | |
| DTCoTBackbone Model=Llama-3.2-3B-Instruct2025.09 | 8.62 | 92.15 | 8.94 | |
| BadChainBackbone Model=Llama-3.2-3B-Instruct2025.09 | 8.18 | 91.82 | 0 |