Faithfulness Detection on LatentAudit Qwen-2.5-7B (evaluation)
94.5AUROCGPT-4o Judge
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GPT-4o JudgeLatency (ms)=∼5,3002026.04 | 94.5 | 87.6 | |
| LatentAuditLatency (ms)=0.77 (+11.2 proving)2026.04 | 93.8 | 86.2 | |
| INSIDELatency (ms)=∼3.82026.04 | 90.1 | 83.2 | |
| SAPLMALatency (ms)=∼1.52026.04 | 87.6 | 80.8 | |
| SelfCheckGPTLatency (ms)=∼28,5002026.04 | 86.5 | 79.8 | |
| Min-PerplexityLatency (ms)=0.02026.04 | 71.8 | 65 |