Uncertainty Quantification on MATH500
65.2ROC-AUC (Threshold 128)LLaMA perplexity
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| LLaMA perplexityBackbone=LLaDA-1.5-8B, Evaluation Model=LLaMA3-Instruct-8B2026.03 | 65.2 | 58.8 | 55 | |
| DiSEBackbone=LLaDA-Instruct-8B2026.03 | 61.1 | 63.4 | 60.4 | |
| DiSEBackbone=LLaDA-1.5-8B2026.03 | 60.6 | 55.3 | 53.3 | |
| MCBackbone=LLaDA-1.5-8B, Nmc=322026.03 | 58 | 54.6 | 55.1 | |
| LLaMA perplexityBackbone=LLaDA-Instruct-8B, Evaluation Model=LLaMA3-Instruct-8B2026.03 | 57.5 | 63.7 | 55.1 | |
| MCBackbone=LLaDA-1.5-8B, Nmc=12026.03 | 55.8 | 51.4 | 52.5 | |
| MCBackbone=LLaDA-Instruct-8B, Nmc=322026.03 | 52.8 | 57.8 | 53.1 | |
| MCBackbone=LLaDA-Instruct-8B, Nmc=12026.03 | 49.7 | 54.1 | 53.2 |