Automated Short Answer Grading Calibration on SciEntsBank
0.138Brier Scorehybrid confidence
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| hybrid confidencealeatoric=with2026.04 | 0.138 | 0.044 | 0.1 | |
| hybrid confidencealeatoric=without2026.04 | 0.172 | 0.066 | 0.155 | |
| consistency-based confidence2026.04 | 0.186 | 0.029 | 0.279 | |
| verbalizing-based confidence2026.04 | 0.194 | 0.051 | 0.259 | |
| latent-based confidence2026.04 | 0.218 | 0.096 | 0.221 |