Data Contamination Detection on SAT
79F1 ScoreEntropy-Noise
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Entropy-NoiseTarget Model=Qwen2.5-7B-Instruct2025.10 | 79 | 77 | |
| Self-CritiqueTarget Model=Qwen2.5-7B-Instruct2025.10 | 69 | 67 | |
| Min-K%Target Model=DeepSeek-Math-7B-Instruct2025.10 | 69 | 35 | |
| RecallTarget Model=DeepSeek-Math-7B-Instruct2025.10 | 69 | 62 | |
| PPLTarget Model=DeepSeek-Math-7B-Instruct2025.10 | 68 | 64 | |
| Entropy-NoiseTarget Model=DeepSeek-Math-7B-Instruct2025.10 | 67 | 45 | |
| Entropy-TempTarget Model=Qwen2.5-7B-Instruct2025.10 | 66 | 69 | |
| CDDTarget Model=DeepSeek-Math-7B-Instruct2025.10 | 66 | 50 | |
| Self-CritiqueTarget Model=DeepSeek-Math-7B-Instruct2025.10 | 66 | 67 | |
| Entropy-TempTarget Model=DeepSeek-Math-7B-Instruct2025.10 | 64 | 61 | |
| RecallTarget Model=Qwen2.5-7B-Instruct2025.10 | 62 | 62 | |
| Min-K%++Target Model=DeepSeek-Math-7B-Instruct2025.10 | 62 | 49 | |
| CDDTarget Model=Qwen2.5-7B-Instruct2025.10 | 57 | 47 | |
| PPLTarget Model=Qwen2.5-7B-Instruct2025.10 | 54 | 50 | |
| Min-K%Target Model=Qwen2.5-7B-Instruct2025.10 | 32 | 50 | |
| Min-K%++Target Model=Qwen2.5-7B-Instruct2025.10 | 0 | 31 |