Factuality Detection on TriviaQA, HotpotQA, and CSQA
72.9Average AUROCDynHD
Evaluation Results
| Method | Links | |
|---|---|---|
| DynHDMethod Category=Trajectory-based, Evaluation Protocol=Zero-shot cross-task generalization2026.03 | 72.9 | |
| TraceDetMethod Category=Trajectory-based, Evaluation Protocol=Zero-shot cross-task generalization2026.03 | 66.3 | |
| TSVMethod Category=Latent-based, Evaluation Protocol=Zero-shot cross-task generalization2026.03 | 61.3 | |
| CCSMethod Category=Latent-based, Evaluation Protocol=Zero-shot cross-task generalization2026.03 | 58.2 |