Unsupervised Error Diagnosis on ARC-C
0.804AUROCBMC
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| BMCModel=Dream-v0-Instruct-7B2026.04 | 0.804 | 0.951 | |
| BMCModel=LLaDA-8B-Instruct2026.04 | 0.777 | 0.918 | |
| Self-ConsistencyModel=LLaDA-8B-Instruct2026.04 | 0.735 | 0.895 | |
| Self-ConsistencyModel=Dream-v0-Instruct-7B2026.04 | 0.708 | 0.922 | |
| Self-EvaluationModel=Dream-v0-Instruct-7B2026.04 | 0.57 | 0.886 | |
| Model ConfidenceModel=LLaDA-8B-Instruct2026.04 | 0.55 | 0.824 | |
| Self-EvaluationModel=LLaDA-8B-Instruct2026.04 | 0.546 | 0.821 | |
| Model ConfidenceModel=Dream-v0-Instruct-7B2026.04 | 0.463 | 0.855 |