Faithfulness Detection on LatentAudit Mistral-7B (evaluation)
94AUROCGPT-4o Judge
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GPT-4o JudgeLatency (ms)=∼5,3002026.04 | 94 | 87 | |
| LatentAuditLatency (ms)=0.77 (+11.2 proving)2026.04 | 92.5 | 85.2 | |
| INSIDELatency (ms)=∼3.82026.04 | 89.5 | 82.5 | |
| SAPLMALatency (ms)=∼1.52026.04 | 87 | 80 | |
| SelfCheckGPTLatency (ms)=∼28,5002026.04 | 85.8 | 79 | |
| Min-PerplexityLatency (ms)=0.02026.04 | 71 | 64.2 |