Hallucination Detection on SQuAD (AUC, F1)
85.5AUCQAODQ⊕V⊥
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| QAODQ⊕V⊥Model=Qwen3-30B-A3B2026.05 | 85.5 | 40.81 | |
| QAODQ⊕V⊥Model=Qwen3-14B2026.05 | 85.01 | 32.09 | |
| SAPLMAModel=Qwen3-14B2026.05 | 83.57 | 29.03 | |
| SAPLMAModel=Qwen3-30B-A3B2026.05 | 83.31 | 39.76 | |
| QAODQ⊕V⊥Model=Llama-2-7B-chat2026.05 | 83.09 | 47.47 | |
| QAODV⊥Model=Qwen3-30B-A3B2026.05 | 82.52 | 36.45 | |
| QAODQ⊕V⊥Model=gemma-2-2b2026.05 | 82.16 | 48.84 | |
| QAODV⊥Model=Qwen3-14B2026.05 | 81.34 | 29.39 | |
| QAODV⊥Model=Llama-2-7B-chat2026.05 | 80.75 | 44.08 | |
| QAODV⊥Model=gemma-2-2b2026.05 | 79.73 | 47.55 | |
| SAPLMAModel=Llama-2-7B-chat2026.05 | 79.47 | 43.25 | |
| SINdexModel=gemma-2-2b2026.05 | 78.26 | 42 | |
| Logit EntropyModel=gemma-2-2b2026.05 | 77.35 | 40.54 | |
| SAPLMAModel=gemma-2-2b2026.05 | 77.23 | 42.29 | |
| PPLModel=gemma-2-2b2026.05 | 76.18 | 40.34 | |
| SINdexModel=Llama-2-7B-chat2026.05 | 75.72 | 39.56 | |
| SINdexModel=Qwen3-30B-A3B2026.05 | 74.07 | 35.46 | |
| SINdexModel=Qwen3-14B2026.05 | 74.02 | 20.46 | |
| MINDModel=Qwen3-30B-A3B2026.05 | 72.9 | 30.48 | |
| MINDModel=Qwen3-14B2026.05 | 72.12 | 24.03 | |
| PikModel=Qwen3-30B-A3B2026.05 | 71.6 | 27.71 | |
| PfalseModel=Qwen3-30B-A3B2026.05 | 70.59 | 22.79 | |
| PikModel=Qwen3-14B2026.05 | 69.76 | 19.45 | |
| PfalseModel=Llama-2-7B-chat2026.05 | 69.42 | 32 | |
| PikModel=gemma-2-2b2026.05 | 69.31 | 35.16 | |
| MINDModel=Llama-2-7B-chat2026.05 | 68.06 | 36.1 | |
| MINDModel=gemma-2-2b2026.05 | 66.76 | 34.2 | |
| PikModel=Llama-2-7B-chat2026.05 | 65.93 | 37.26 | |
| PfalseModel=gemma-2-2b2026.05 | 65.29 | 33.65 | |
| LLM-CheckModel=Qwen3-14B2026.05 | 62.64 | 13.06 | |
| LLM-CheckModel=gemma-2-2b2026.05 | 59.18 | 28.13 | |
| LLM-CheckModel=Qwen3-30B-A3B2026.05 | 58.69 | 18.21 | |
| PfalseModel=Qwen3-14B2026.05 | 58.33 | 12.33 | |
| LLM-CheckModel=Llama-2-7B-chat2026.05 | 57.35 | 25.5 | |
| Logit EntropyModel=Llama-2-7B-chat2026.05 | 56.48 | 22.59 | |
| PPLModel=Llama-2-7B-chat2026.05 | 56.19 | 22.59 | |
| Logit EntropyModel=Qwen3-30B-A3B2026.05 | 55.82 | 17.74 | |
| PPLModel=Qwen3-30B-A3B2026.05 | 55.78 | 17.74 | |
| Logit EntropyModel=Qwen3-14B2026.05 | 54.92 | 14.65 | |
| PPLModel=Qwen3-14B2026.05 | 54.89 | 14.75 |