Hallucination Detection on RAGTruth RT-QA 1.0 (test)
0.7885F1 ScoreFourier-high
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Fourier-highModel=Mistral-7B2026.02 | 0.7885 | 0.9099 | |
| Wavelet-highModel=Mistral-7B2026.02 | 0.7876 | 0.9117 | |
| Lookback-lensModel=Mistral-7B2026.02 | 0.7832 | 0.9148 | |
| Attn-entropyModel=Mistral-7B2026.02 | 0.781 | 0.9083 | |
| LaplacianModel=Mistral-7B2026.02 | 0.7807 | 0.9049 | |
| Fourier-highModel=LLaMA-7B2026.02 | 0.7277 | 0.8584 | |
| Wavelet-highModel=LLaMA-7B2026.02 | 0.7194 | 0.8526 | |
| LaplacianModel=LLaMA-7B2026.02 | 0.7107 | 0.8449 | |
| Fourier-highModel=LLaMA-13B2026.02 | 0.7068 | 0.8792 | |
| Wavelet-highModel=LLaMA-13B2026.02 | 0.7029 | 0.8741 | |
| Lookback-lensModel=LLaMA-13B2026.02 | 0.6947 | 0.8727 | |
| Lookback-lensModel=LLaMA-7B2026.02 | 0.693 | 0.8482 | |
| Attn-entropyModel=LLaMA-13B2026.02 | 0.6909 | 0.865 | |
| LaplacianModel=LLaMA-13B2026.02 | 0.6834 | 0.8552 | |
| Attn-entropyModel=LLaMA-7B2026.02 | 0.6832 | 0.8481 | |
| SelfcheckGPTModel=Mistral-7B2026.02 | 0.6551 | 0.7084 | |
| SelfcheckGPTModel=LLaMA-7B2026.02 | 0.6289 | 0.6942 | |
| RedeepModel=Mistral-7B2026.02 | 0.627 | 0.7357 | |
| RefCheckerModel=Mistral-7B2026.02 | 0.6144 | 0.6121 | |
| SelfcheckGPTModel=LLaMA-13B2026.02 | 0.6029 | 0.6425 | |
| RedeepModel=LLaMA-7B2026.02 | 0.5972 | 0.6364 | |
| RefCheckerModel=LLaMA-13B2026.02 | 0.5928 | 0.5893 | |
| RefCheckerModel=LLaMA-7B2026.02 | 0.5914 | 0.5865 | |
| RedeepModel=LLaMA-13B2026.02 | 0.574 | 0.6457 | |
| Attn-varianceModel=LLaMA-13B2026.02 | 0.5636 | 0.7497 | |
| Attn-varianceModel=Mistral-7B2026.02 | 0.5636 | 0.7497 | |
| EigenScoreModel=LLaMA-7B2026.02 | 0.4979 | 0.5253 | |
| EigenScoreModel=Mistral-7B2026.02 | 0.4904 | 0.4611 | |
| Attn-varianceModel=LLaMA-7B2026.02 | 0.4807 | 0.6147 | |
| EigenScoreModel=LLaMA-13B2026.02 | 0.4788 | 0.4734 | |
| Coverage onlyFeature Set=Coverage only2026.06 | 0.296 | 0.589 | |
| Support density onlyFeature Set=Support density only2026.06 | 0.289 | 0.557 | |
| EGCFeature Set=all features2026.06 | 0.285 | 0.556 |