Hallucination Detection on LLM-AggreFact
91.99CNN PerformanceCPIL-1%
Evaluation Results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| CPIL-1%labeled data fraction=1%, backbone=DeBERTa2026.06 | 91.99 | 73.47 | 83.02 | 86.13 | 69.2 | 77.13 | 85.1 | 61.46 | 70.57 | 87.85 | 84.74 | 79.14 | |
| CPIL-2%labeled data fraction=2%, backbone=DeBERTa2026.06 | 91.08 | 71.38 | 84.05 | 90.1 | 64.91 | 77.77 | 81.74 | 58.36 | 71.94 | 88.9 | 84.79 | 78.64 | |
| CCHD-FT5Backbone=Flan-T52026.06 | 88.58 | 73.37 | 78.92 | 88.34 | 68.31 | 76.62 | 86.5 | 62.46 | 72.41 | 88.28 | 86.42 | 79.11 | |
| CCHD-DBTBackbone=DeBERTa2026.06 | 88.05 | 77.1 | 84.74 | 87.43 | 70.7 | 78.29 | 83.81 | 61.47 | 68.53 | 89.46 | 87.4 | 79.73 | |
| AlignScore2026.06 | 81.71 | 68.51 | 80.59 | 85.18 | 54.75 | 69.79 | 81.16 | 61.05 | 70.63 | 85.81 | 77.35 | 74.23 | |
| AlignScore2026.06 | 81.71 | 68.51 | 80.59 | 85.18 | 54.75 | 69.79 | 81.16 | 61.05 | 70.63 | 85.81 | 77.35 | 74.23 | |
| FactCG-FT5Backbone=Flan-T52026.06 | 81.42 | 74.52 | 82.36 | 88.74 | 68.18 | 75.05 | 87.05 | 59.57 | 70.73 | 88.7 | 86.82 | 78.47 | |
| FactCG2026.06 | 78.44 | 74.46 | 83.98 | 80.07 | 69.87 | 80.53 | 83.45 | 63.42 | 58.54 | 89.12 | 86.63 | 77.14 | |
| FactCG-DBTBackbone=DeBERTa2026.06 | 78.44 | 74.46 | 83.98 | 80.07 | 69.87 | 80.53 | 83.45 | 63.42 | 58.54 | 89.12 | 86.63 | 77.14 | |
| FactCC2026.06 | 78.3 | 69.2 | 82.46 | 86.74 | 69.9 | 77.77 | 86.41 | 61.28 | 65.15 | 87.47 | 86.1 | 77.34 | |
| MiniCheck2026.06 | 74.88 | 74.49 | 83.72 | 81.26 | 62.1 | 72.31 | 85.54 | 60.15 | 72.22 | 88.15 | 86.36 | 76.47 | |
| MiniCheck-DBTBackbone=DeBERTa2026.06 | 74.88 | 74.49 | 83.72 | 81.26 | 62.1 | 72.31 | 85.54 | 60.15 | 72.22 | 88.15 | 86.36 | 76.47 | |
| MiniCheck-FT5Backbone=Flan-T52026.06 | 72.64 | 71.22 | 76.2 | 89.49 | 65.97 | 74.16 | 81.88 | 63.39 | 71.99 | 89.06 | 84.28 | 76.39 | |
| SummaC-CVmode=Cross-Validation2026.06 | 66.41 | 48.56 | 69.7 | 77.87 | 50.98 | 70.77 | 68.36 | 57.53 | 65.21 | 79.16 | 56.92 | 64.68 | |
| SummaC-CVEvaluation Protocol=cross-validation2026.06 | 66.41 | 48.56 | 69.7 | 77.87 | 50.98 | 70.77 | 68.36 | 57.53 | 65.21 | 79.16 | 56.92 | 64.68 | |
| SummaC-ZSmode=Zero-Shot2026.06 | 55.19 | 66.55 | 79.57 | 80.14 | 41.03 | 74.5 | 72.96 | 60.79 | 65.3 | 81.58 | 62.81 | 67.31 | |
| SummaC-ZSEvaluation Protocol=zero-shot2026.06 | 55.19 | 66.55 | 79.57 | 80.14 | 41.03 | 74.5 | 72.96 | 60.79 | 65.3 | 81.58 | 62.81 | 67.31 |