Confusion-Aware Rubric Optimization for LLM-based Automated Grading

TL;DR

This paper introduces CARO, a novel framework that improves LLM-based grading accuracy by structurally decomposing error signals using confusion matrices, enabling targeted fixes and avoiding rule dilution.

Contribution

CARO is the first approach to decompose error signals into modes for targeted rubric optimization, enhancing accuracy and efficiency in automated grading.

Findings

CARO outperforms state-of-the-art methods on education datasets.

Structural error decomposition improves grading precision.

Targeted mode-specific repairs enhance scalability and robustness.

Abstract

Accurate and unambiguous guidelines are critical for large language model (LLM) based graders, yet manually crafting these prompts is often sub-optimal as LLMs can misinterpret expert guidelines or lack necessary domain specificity. Consequently, the field has moved toward automated prompt optimization to refine grading guidelines without the burden of manual trial and error. However, existing frameworks typically aggregate independent and unstructured error samples into a single update step, resulting in "rule dilution" where conflicting constraints weaken the model's grading logic. To address these limitations, we introduce Confusion-Aware Rubric Optimization (CARO), a novel framework that enhances accuracy and computational efficiency by structurally separating error signals. CARO leverages the confusion matrix to decompose monolithic error signals into distinct modes, allowing for the diagnosis and repair of specific misclassification patterns individually. By synthesizing targeted "fixing patches" for dominant error modes and employing a diversity-aware selection mechanism, the framework prevents guidance conflict and eliminates the need for resource-heavy nested refinement loops. Empirical evaluations on teacher education and STEM datasets demonstrate that CARO significantly outperforms existing SOTA methods. These results suggest that replacing mixed-error aggregation with surgical, mode-specific repair yields robust improvements in automated assessment scalability and precision.