Hallucination self-detection on WikiBio GPT-4o
85AccuracySelf-Confidence
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Self-ConfidenceModel=GPT-4o, Sampling Strategy=Single-Sample, Knowledge Graph Support=Enabled2025.12 | 85 | 87 | 88 | |
| Self-ConfidenceModel=Gemini-2.5-Flash, Sampling Strategy=Single-Sample, Knowledge Graph Support=Enabled2025.12 | 84 | 85 | 87 | |
| Self-QuestioningModel=GPT-4o, Sampling Strategy=Single-Sample, Knowledge Graph Support=Enabled2025.12 | 79 | 81 | 84 | |
| Self-ConfidenceModel=GPT-4o, Sampling Strategy=Single-Sample, Knowledge Graph Support=Disabled2025.12 | 79 | 81 | 86 | |
| Self-ConfidenceModel=Gemini-2.5-Flash, Sampling Strategy=Single-Sample, Knowledge Graph Support=Disabled2025.12 | 78 | 79 | 83 | |
| Self-QuestioningModel=GPT-4o, Sampling Strategy=Single-Sample, Knowledge Graph Support=Disabled2025.12 | 72 | 71 | 78 | |
| Self-Check GPTModel=GPT-4o, Sampling Strategy=Multi-Sample, Knowledge Graph Support=Enabled2025.12 | 72 | 76 | 75 | |
| Self-Check GPTModel=Gemini-2.5-Flash, Sampling Strategy=Multi-Sample, Knowledge Graph Support=Enabled2025.12 | 70 | 74 | 73 | |
| Self-QuestioningModel=Gemini-2.5-Flash, Sampling Strategy=Single-Sample, Knowledge Graph Support=Enabled2025.12 | 69 | 73 | 73 | |
| Self-Check GPTModel=GPT-4o, Sampling Strategy=Multi-Sample, Knowledge Graph Support=Disabled2025.12 | 63 | 69 | 62 | |
| Self-QuestioningModel=Gemini-2.5-Flash, Sampling Strategy=Single-Sample, Knowledge Graph Support=Disabled2025.12 | 62 | 63 | 63 | |
| Self-Check GPTModel=Gemini-2.5-Flash, Sampling Strategy=Multi-Sample, Knowledge Graph Support=Disabled2025.12 | 60 | 68 | 55 |