Hallucination Detection on SciQ (AUROC, F1-Score)
0.9328AUROCCausalGaze
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CausalGazeLLM=Qwen2-7B2026.04 | 0.9328 | 82.83 | |
| HaluGNNLLM=Qwen2-7B2026.04 | 0.9217 | 84.02 | |
| CausalGazeLLM=Mistral-7B2026.04 | 0.855 | 78.6 | |
| CausalGazeLLM=Llama2-7B2026.04 | 0.839 | 77.01 | |
| HaluGNNLLM=Mistral-7B2026.04 | 0.8343 | 72.73 | |
| HaluGNNLLM=Llama2-7B2026.04 | 0.8336 | 76.49 | |
| SAPLMALLM=Qwen2-7B2026.04 | 0.815 | 82 | |
| SAPLMALLM=Mistral-7B2026.04 | 0.7884 | 77.9 | |
| ICR-ProbeLLM=Mistral-7B2026.04 | 0.7793 | 78.26 | |
| ICR-ProbeLLM=Llama2-7B2026.04 | 0.7748 | 75.61 | |
| SAPLMALLM=Llama2-7B2026.04 | 0.773 | 77.67 | |
| ICR-ProbeLLM=Qwen2-7B2026.04 | 0.7595 | 72.15 | |
| Semantic EntropyLLM=Mistral-7B2026.04 | 0.6945 | 66.4 | |
| Semantic EntropyLLM=Llama2-7B2026.04 | 0.682 | 67.97 | |
| SelfCheckGPTLLM=Llama2-7B2026.04 | 0.679 | 68.3 | |
| Semantic EntropyLLM=Qwen2-7B2026.04 | 0.659 | 66.04 | |
| EigenScoreLLM=Mistral-7B2026.04 | 0.6474 | 63.46 | |
| EigenScoreLLM=Qwen2-7B2026.04 | 0.632 | 64.62 | |
| LN-EntropyLLM=Mistral-7B2026.04 | 0.6036 | 58.32 | |
| EigenScoreLLM=Llama2-7B2026.04 | 0.596 | 60.08 | |
| SelfCheckGPTLLM=Qwen2-7B2026.04 | 0.586 | 57.79 | |
| LLM-CheckLLM=Llama2-7B2026.04 | 0.5842 | 56.37 | |
| LN-EntropyLLM=Llama2-7B2026.04 | 0.576 | 58.26 | |
| LLM-CheckLLM=Mistral-7B2026.04 | 0.5748 | 55.47 | |
| LLM-CheckLLM=Qwen2-7B2026.04 | 0.5726 | 57.73 | |
| P(True)LLM=Mistral-7B2026.04 | 0.5673 | 55.31 | |
| SelfCheckGPTLLM=Mistral-7B2026.04 | 0.5593 | 52.6 | |
| PerplexityLLM=Mistral-7B2026.04 | 0.547 | 55.83 | |
| P(True)LLM=Llama2-7B2026.04 | 0.546 | 56.45 | |
| P(True)LLM=Qwen2-7B2026.04 | 0.538 | 54.49 | |
| PerplexityLLM=Qwen2-7B2026.04 | 0.534 | 54.77 | |
| PerplexityLLM=Llama2-7B2026.04 | 0.526 | 52.56 | |
| LN-EntropyLLM=Qwen2-7B2026.04 | 0.524 | 52.56 |