Text Anonymization on MedQA
24.6Privacy ScoreOptimized Prompt
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Optimized PromptBackbone Model=Qwen3-30B-A3B2026.02 | 24.6 | 45.9 | |
| Qwen3-30B-A3B (Optimized)Backbone=Qwen3-30B-A3B, Status=Optimized2026.02 | 24.6 | 45.9 | |
| AFBackbone=GPT-52026.02 | 24.4 | 45.8 | |
| Optimized PromptBackbone Model=Gemma-3-27b-it2026.02 | 14.3 | 56.8 | |
| Seed PromptBackbone Model=Gemma-3-27b-it2026.02 | 12 | 46.8 | |
| AFBackbone Model=Gemma-3-27b-it2026.02 | 11.7 | 29.4 | |
| Task-Specific PromptBackbone Model=Gemma-3-27b-it2026.02 | 10.7 | 35.3 | |
| PromptBackbone=GPT-52026.02 | 10.7 | 45.5 | |
| AFBackbone Model=Mistral-Small-3.2-24B2026.02 | 9.39 | 41.2 | |
| Task-Specific PromptBackbone Model=Qwen3-30B-A3B2026.02 | 8.25 | 41.2 | |
| Optimized PromptBackbone Model=Mistral-Small-3.2-24B2026.02 | 8.06 | 53.2 | |
| Task-Specific PromptBackbone Model=Mistral-Small-3.2-24B2026.02 | 7.33 | 35.3 | |
| AFBackbone Model=Qwen3-30B-A3B2026.02 | 6.53 | 29.4 | |
| OpenPII2026.02 | 3.8 | 59.5 | |
| Seed PromptBackbone Model=Qwen3-30B-A3B2026.02 | 3.52 | 58.6 | |
| Seed PromptBackbone Model=Mistral-Small-3.2-24B2026.02 | 2.91 | 54.9 |