CHiL(L)Grader: Calibrated Human-in-the-Loop Short-Answer Grading
This addresses the need for safe, high-stakes educational assessment by providing a calibrated system that adapts to evolving curricula, though it is incremental in combining existing techniques like temperature scaling and selective prediction.
The paper tackled the problem of unreliable confidence in large language models for automated short-answer grading by introducing CHiL(L)Grader, a human-in-the-loop framework with calibrated confidence estimation, which automatically scored 35-65% of responses at expert-level quality (QWK >= 0.80) across three datasets.
Scaling educational assessment with large language models requires not just accuracy, but the ability to recognize when predictions are trustworthy. Instruction-tuned models tend to be overconfident, and their reliability deteriorates as curricula evolve, making fully autonomous deployment unsafe in high-stakes settings. We introduce CHiL(L)Grader, the first automated grading framework that incorporates calibrated confidence estimation into a human-in-the-loop workflow. Using post-hoc temperature scaling, confidence-based selective prediction, and continual learning, CHiL(L)Grader automates only high-confidence predictions while routing uncertain cases to human graders, and adapts to evolving rubrics and unseen questions. Across three short-answer grading datasets, CHiL(L)Grader automatically scores 35-65% of responses at expert-level quality (QWK >= 0.80). A QWK gap of 0.347 between accepted and rejected predictions confirms the effectiveness of the confidence-based routing. Each correction cycle strengthens the model's grading capability as it learns from teacher feedback. These results show that uncertainty quantification is key for reliable AI-assisted grading.