PII Detection on MathEd-PII (test)
93.4PrecisionClaude 4.5 Opus
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Claude 4.5 OpusPrompting Strategy=Segment-Aware Prompting2026.02 | 93.4 | 73 | 82 | |
| Claude 4.5 OpusPrompting Strategy=Basic Prompting2026.02 | 91.1 | 73 | 81 | |
| Claude 4.5 OpusPrompting Strategy=Math-Aware Prompting2026.02 | 90.5 | 72 | 80.2 | |
| Gemini 3 ProPrompting Strategy=Segment-Aware Prompting2026.02 | 88.8 | 76.4 | 82.1 | |
| Gemini 3 ProPrompting Strategy=Basic Prompting2026.02 | 88.7 | 76 | 81.8 | |
| Gemini 3 ProPrompting Strategy=Math-Aware Prompting2026.02 | 86.9 | 73.5 | 79.6 | |
| Gemini 2.5 FlashPrompting Strategy=Basic Prompting2026.02 | 84.8 | 77.1 | 80.7 | |
| Claude 4.5 HaikuPrompting Strategy=Basic Prompting2026.02 | 83.1 | 62 | 71 | |
| GPT-5.2Prompting Strategy=Segment-Aware Prompting2026.02 | 80 | 72 | 75.8 | |
| GPT-5.2Prompting Strategy=Math-Aware Prompting2026.02 | 77.4 | 72.1 | 74.7 | |
| GPT-5.2Prompting Strategy=Basic Prompting2026.02 | 75.9 | 70.7 | 73.2 | |
| GPT-4 TurboPrompting Strategy=Basic Prompting2026.02 | 73.2 | 57.8 | 64.6 | |
| Presidio (Large)Prompting Strategy=Baseline2026.02 | 25.4 | 74.7 | 37.9 | |
| Presidio (Transformer)Prompting Strategy=Baseline2026.02 | 23 | 78.1 | 35.5 |