Data Contamination Detection on MATH-500 (test)
42.6Reference Score SGPT-5.1
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| GPT-5.1Metric=Con(%), Detection Method=ZCP2026.05 | 42.6 | 44.4 | 52.1 | |
| GPT-5.1Metric=ACC(%), Detection Method=ZCP2026.05 | 37.6 | 39.6 | 54.9 | |
| GPT-4oMetric=ACC(%), Detection Method=ZCP2026.05 | 32.93 | 38 | 98.4 | |
| GPT-4oMetric=Con(%), Detection Method=ZCP2026.05 | 32.93 | 39.8 | 98.9 | |
| Qwen-2.5-MathMetric=ACC(%), Detection Method=ZCP2026.05 | 32.4 | 34 | 52 | |
| Qwen-2.5-MathMetric=Con(%), Detection Method=ZCP2026.05 | 32.4 | 33.4 | 50.1 | |
| Qwen-3Metric=ACC(%), Detection Method=ZCP2026.05 | 26 | 29.4 | 71.5 | |
| Qwen-3Metric=Con(%), Detection Method=ZCP2026.05 | 26 | 27.4 | 51.4 | |
| Qwen-2.5-MathMetric=P_first, Detection Method=ZCP2026.05 | 0.269 | 0.356 | 99.8 | |
| Qwen-2.5-MathMetric=P_all, Detection Method=ZCP2026.05 | 0.256 | 0.33 | 99.8 | |
| Qwen-3Metric=P_first, Detection Method=ZCP2026.05 | 0.251 | 0.339 | 99.8 | |
| Qwen-3Metric=P_all, Detection Method=ZCP2026.05 | 0.192 | 0.275 | 99.8 |