Unsupervised Error Diagnosis on MATH
82.5AUROCBMC
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| BMCModel=Dream-v0-Instruct-7B2026.04 | 82.5 | 66 | |
| BMCModel=LLaDA-8B-Instruct2026.04 | 82 | 61.3 | |
| Self-ConsistencyModel=LLaDA-8B-Instruct2026.04 | 80.3 | 54.6 | |
| Model ConfidenceModel=LLaDA-8B-Instruct2026.04 | 71.3 | 43 | |
| Self-ConsistencyModel=Dream-v0-Instruct-7B2026.04 | 67.5 | 40.8 | |
| Self-EvaluationModel=LLaDA-8B-Instruct2026.04 | 55.8 | 26.6 | |
| Self-EvaluationModel=Dream-v0-Instruct-7B2026.04 | 52.5 | 24.5 | |
| Model ConfidenceModel=Dream-v0-Instruct-7B2026.04 | 52.2 | 22.8 |