Factual Consistency Detection on TofuEval-MeetingBank Response level leakage-clean protocol (test)
65.9AUCGASP+base (trained)
Evaluation Results
| Method | Links | |
|---|---|---|
| GASP+base (trained)Model=Qwen-0.5B2026.07 | 65.9 | |
| GASP+base (trained)Model=SmolLM22026.07 | 65.6 | |
| GASP-trainedModel=SmolLM22026.07 | 64.1 | |
| GASP+base (trained)Model=Qwen-1.5B2026.07 | 63.3 | |
| GASP-trainedModel=Qwen-0.5B2026.07 | 63.2 | |
| GASP-trainedModel=Qwen-1.5B2026.07 | 62.7 | |
| GASP-threshold (default)Model=Qwen-0.5B2026.07 | 62.5 | |
| GASP-threshold (default)Model=Qwen-1.5B2026.07 | 62.3 | |
| GASP-threshold (default)Model=SmolLM22026.07 | 61.4 | |
| PerplexityModel=Qwen-1.5B2026.07 | 54.8 | |
| PerplexityModel=Qwen-0.5B2026.07 | 52.1 | |
| PerplexityModel=SmolLM22026.07 | 51.6 |