CHiL(L)Grader: Calibrated Human-in-the-Loop Short-Answer Grading

TL;DR

This paper presents CHiL(L)Grader, an AI-assisted grading system that uses calibrated confidence estimates to automate high-confidence responses and involve humans for uncertain cases, improving reliability and adaptability in educational assessments.

Contribution

It introduces a novel framework combining calibrated confidence estimation with human-in-the-loop workflows for scalable, reliable short-answer grading.

Findings

Automates 35-65% of responses at expert-level quality

Confidence-based routing significantly reduces grading errors

Model improves over correction cycles through continual learning

Abstract

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.