Privacy Violation Detection on CMPL Insurance Single-turn (test)
0rbypassNeuroFilter
Evaluation results
| Method | Links | ||
|---|---|---|---|
| NeuroFilterModel Scale=7B Model, Setting=Single-turn2026.01 | 0 | 0 | |
| SAE-basedModel Scale=7B Model, Setting=Single-turn2026.01 | 6 | 96 | |
| Llama GuardModel Scale=7B Model, Setting=Single-turn2026.01 | 100 | 0 |