Hallucination Detection on 6292 OR-expert-labeled artifacts natural benchmark
61.7Hallucination F1OptArgus
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| OptArgusBackbone=DeepSeek-V3.22026.05 | 61.7 | 54.1 | 49.5 | 57.9 | 43.1 | 51.2 | 48.6 | |
| Single-Agent DetectorBackbone=DeepSeek-V3.22026.05 | 52.1 | 40.4 | 32.8 | 50.4 | 29.4 | 38.2 | 37.9 |