Hallucination Detection on PsiloQA
92.64AUROCCurvature
Evaluation Results
| Method | Links | |
|---|---|---|
| CurvatureBackbone=Mistral-7B-Instruct-v0.3, First-sentence truncation=true, Forward horizon (w)=72026.05 | 92.64 | |
| Val LossBackbone=LLaMA-3.1-8B-Instruct, First-sentence truncation=true, Forward horizon (w)=72026.05 | 92.38 | |
| Val LossBackbone=Mistral-7B-Instruct-v0.3, First-sentence truncation=true, Forward horizon (w)=72026.05 | 92.16 | |
| IDBackbone=LLaMA-3.1-8B-Instruct, First-sentence truncation=true, Forward horizon (w)=72026.05 | 92.08 | |
| FEPoIDBackbone=LLaMA-3.1-8B-Instruct, First-sentence truncation=true, Forward horizon (w)=72026.05 | 92.08 | |
| FEPoIDBackbone=Mistral-7B-Instruct-v0.3, First-sentence truncation=true, Forward horizon (w)=72026.05 | 92.07 | |
| CurvatureBackbone=LLaMA-3.1-8B-Instruct, First-sentence truncation=true, Forward horizon (w)=72026.05 | 91.6 | |
| Val LossBackbone=LlaMA-3.1-8B, Forward horizon (w)=7, Representation extraction method=FST, Model tuning status=non-instruction-tuned (base model)2026.05 | 90.51 | |
| CurvatureBackbone=LlaMA-3.1-8B, Forward horizon (w)=7, Representation extraction method=FST, Model tuning status=non-instruction-tuned (base model)2026.05 | 90.49 | |
| FEPoIDLLM Backbone=Mistral-7B-Instruct-v0.3, Method Category=Hidden-State Probing2026.05 | 90.31 | |
| IDBackbone=LlaMA-3.1-8B, Forward horizon (w)=7, Representation extraction method=FST, Model tuning status=non-instruction-tuned (base model)2026.05 | 89.85 | |
| FEPoIDBackbone=LlaMA-3.1-8B, Forward horizon (w)=7, Representation extraction method=FST, Model tuning status=non-instruction-tuned (base model)2026.05 | 89.85 | |
| IDBackbone=Mistral-7B-Instruct-v0.3, First-sentence truncation=true, Forward horizon (w)=72026.05 | 89.42 | |
| RGNBackbone=Mistral-7B-Instruct-v0.3, First-sentence truncation=true, Forward horizon (w)=72026.05 | 87.97 | |
| IDBackbone=LlaMA-3.1-8B base, Forward horizon (w)=7, FST=false2026.05 | 86.55 | |
| FEPoIDBackbone=LlaMA-3.1-8B base, Forward horizon (w)=7, FST=false2026.05 | 86.55 | |
| CurvatureBackbone=LlaMA-3.1-8B base, Forward horizon (w)=7, FST=false2026.05 | 86.39 | |
| RankMEBackbone=Mistral-7B-Instruct-v0.3, First-sentence truncation=true, Forward horizon (w)=72026.05 | 86.17 | |
| RGNBackbone=LlaMA-3.1-8B, Forward horizon (w)=7, Representation extraction method=FST, Model tuning status=non-instruction-tuned (base model)2026.05 | 85.83 | |
| RankMEBackbone=LLaMA-3.1-8B-Instruct, First-sentence truncation=true, Forward horizon (w)=72026.05 | 85.49 | |
| Val LossBackbone=LlaMA-3.2-3B, Forward horizon (w)=7, FST extraction status=without FST2026.05 | 84.98 | |
| FEPoIDBackbone=LlaMA-3.2-3B, Forward horizon (w)=7, FST extraction status=without FST2026.05 | 84.83 | |
| Val LossBackbone=LlaMA-3.1-8B base, Forward horizon (w)=7, FST=false2026.05 | 84.34 | |
| RankMEBackbone=LlaMA-3.1-8B, Forward horizon (w)=7, Representation extraction method=FST, Model tuning status=non-instruction-tuned (base model)2026.05 | 83.98 | |
| CurvatureBackbone=LlaMA-3.2-3B, Forward horizon (w)=7, FST extraction status=without FST2026.05 | 82.8 | |
| SNRBackbone=Mistral-7B-Instruct-v0.3, First-sentence truncation=true, Forward horizon (w)=72026.05 | 82.74 | |
| RGNBackbone=LLaMA-3.1-8B-Instruct, First-sentence truncation=true, Forward horizon (w)=72026.05 | 81.12 | |
| Lexical SimilarityBackbone=LLaMA-3.1-8B-Instruct, First-sentence truncation=true, Forward horizon (w)=72026.05 | 80.61 | |
| RankMEBackbone=LlaMA-3.1-8B base, Forward horizon (w)=7, FST=false2026.05 | 80.32 | |
| SNRBackbone=LLaMA-3.1-8B-Instruct, First-sentence truncation=true, Forward horizon (w)=72026.05 | 80.01 | |
| SNRBackbone=LlaMA-3.1-8B, Forward horizon (w)=7, Representation extraction method=FST, Model tuning status=non-instruction-tuned (base model)2026.05 | 79.99 | |
| RGNBackbone=LlaMA-3.2-3B, Forward horizon (w)=7, FST extraction status=without FST2026.05 | 79.9 | |
| IDBackbone=LlaMA-3.2-3B, Forward horizon (w)=7, FST extraction status=without FST2026.05 | 79.9 | |
| EigenScoreBackbone=LLaMA-3.1-8B-Instruct, First-sentence truncation=true, Forward horizon (w)=72026.05 | 79.41 | |
| Pred. EntropyBackbone=LLaMA-3.1-8B-Instruct, First-sentence truncation=true, Forward horizon (w)=72026.05 | 78.8 | |
| FEPoIDLLM Backbone=LlaMA-3.1-8B-Instruct, Method Category=Hidden-State Probing2026.05 | 78.62 | |
| RGNBackbone=LlaMA-3.1-8B base, Forward horizon (w)=7, FST=false2026.05 | 78.16 | |
| SNRBackbone=LlaMA-3.1-8B base, Forward horizon (w)=7, FST=false2026.05 | 76.77 | |
| Val LossBackbone=LlaMA-3.2-1B, Forward horizon (w)=3, FST extraction status=without FST2026.05 | 75.97 | |
| Pred. EntropyBackbone=Mistral-7B-Instruct-v0.3, First-sentence truncation=true, Forward horizon (w)=72026.05 | 75.74 | |
| EigenScoreBackbone=Mistral-7B-Instruct-v0.3, First-sentence truncation=true, Forward horizon (w)=72026.05 | 75.17 | |
| RankMEBackbone=LlaMA-3.2-3B, Forward horizon (w)=7, FST extraction status=without FST2026.05 | 74.69 | |
| FEPoIDBackbone=LlaMA-3.2-1B, Forward horizon (w)=3, FST extraction status=without FST2026.05 | 74.13 | |
| Semantic EntropyBackbone=LLaMA-3.1-8B-Instruct, First-sentence truncation=true, Forward horizon (w)=72026.05 | 73.92 | |
| CurvatureBackbone=LlaMA-3.2-1B, Forward horizon (w)=3, FST extraction status=without FST2026.05 | 73.55 | |
| Lexical SimilarityBackbone=Mistral-7B-Instruct-v0.3, First-sentence truncation=true, Forward horizon (w)=72026.05 | 69.07 | |
| SNRBackbone=LlaMA-3.2-3B, Forward horizon (w)=7, FST extraction status=without FST2026.05 | 68.78 | |
| Semantic EntropyBackbone=Mistral-7B-Instruct-v0.3, First-sentence truncation=true, Forward horizon (w)=72026.05 | 66.85 | |
| LIDBackbone=LLaMA-3.1-8B-Instruct, First-sentence truncation=true, Forward horizon (w)=72026.05 | 66.48 | |
| LIDBackbone=Mistral-7B-Instruct-v0.3, First-sentence truncation=true, Forward horizon (w)=72026.05 | 66 | |
| RGNBackbone=LlaMA-3.2-1B, Forward horizon (w)=3, FST extraction status=without FST2026.05 | 65.97 | |
| IDBackbone=LlaMA-3.2-1B, Forward horizon (w)=3, FST extraction status=without FST2026.05 | 65.97 | |
| RankMEBackbone=LlaMA-3.2-1B, Forward horizon (w)=3, FST extraction status=without FST2026.05 | 64.41 | |
| SNRBackbone=LlaMA-3.2-1B, Forward horizon (w)=3, FST extraction status=without FST2026.05 | 61.26 | |
| LN-Pred. EntropyBackbone=LLaMA-3.1-8B-Instruct, First-sentence truncation=true, Forward horizon (w)=72026.05 | 57.42 | |
| LN-Pred. EntropyBackbone=Mistral-7B-Instruct-v0.3, First-sentence truncation=true, Forward horizon (w)=72026.05 | 50.45 |