arXiv subjects

Ruochen Sun

Publications and source records attributed to Ruochen Sun.

3 recordsLinked to original sources

arXiv2026

When Rubrics Change: Cross-Rubric Generalization for Critical Thinking Essay Scoring

Automated essay scoring (AES) research has largely focused on cross-prompt generalization, where essays from unseen prompts are scored while the scoring criteria are typically held constant. In practice, however, educators may revise or even introduce new rubrics in their scoring task, to evaluate different aspects of essays. We study cross-rubric generalization: training on essays labeled under one set of rubrics and evaluating on previously unseen rubrics, which target different aspects of the essay. We use a Large Language Model (LLM) fine-tuning framework with two components: rubric-agnostic intermediate representations, called traits, and target-essay supervision under seen rubrics during training. On an AES dataset augmented with multiple rubric-defined labels of student critical thinking skills, we find that traits improve macro F1 by 5.0% over a baseline without traits in the hardest setting, where both target rubrics and target essays are unseen during training. We further find that increasing target-essay supervision improves performance, with our best fine-tuned open-source Llama-based model outperforming GPT-5-mini prompting by 2.1% macro F1 and trailing GPT-5 by 1.9%. These results show that trait-based intermediate structure and controlled supervision improve generalization to unseen rubrics.

cs.CL↗

arXiv2026

CodeGENCAT: Generative Computerized Adaptive Testing for Open-ended Coding Problems

Existing Computerized Adaptive Testing (CAT) frameworks typically select questions based on the predicted likelihood that the student will answer correctly. This design ignores information contained in students' open-ended responses, especially in domains such as programming education, where code structures and bugs contain rich information on student knowledge. In this work, we propose \textbf{Code} \textbf{GEN}erative \textbf{CAT} (\textbf{CodeGENCAT}), a generative CAT framework that selects questions using predicted student code responses. First, we develop a Generative Item Response Theory (GIRT) model that generates code responses conditioned on estimated student knowledge, trained with supervised fine-tuning followed by direct preference optimization for knowledge-response alignment. Second, we introduce three question-selection algorithms that measure uncertainty, coding style diversity, and information from predicted student code responses. Experiments on two real-world programming education datasets show that CodeGENCAT outperforms all CAT baselines, achieving an AUC improvement of up to 4.32\% over the strongest baseline in the early stages of adaptive testing.

cs.CL↗

arXiv2020

A calibration framework for high-resolution hydrological models using a multiresolution and heterogeneous strategy

Increasing spatial and temporal resolution of numerical models continues to propel progress in hydrological sciences, but, at the same time, it has strained the ability of modern automatic calibration methods to produce realistic model parameter combinations for these models. This paper presents a new reliable and fast automatic calibration framework to address this issue. In essence, the proposed framework, adopting a divide and conquer strategy, first partitions the parameters into groups of different resolutions based on their sensitivity or importance, in which the most sensitive parameters are prioritized with highest resolution in parameter search space, while the least sensitive ones are explored with the coarsest resolution at beginning. This is followed by an optimization based iterative calibration procedure consisting of a series of sub-tasks or runs. Between consecutive runs, the setup configuration is heterogeneous with parameter search ranges and resolutions varying among groups. At the completion of each sub-task, the parameter ranges within each group are systematically refined from their previously estimated ranges which are initially based on a priori information. Parameters attain stable convergence progressively with each run. A comparison of this new calibration framework with a traditional optimization-based approach was performed using a quasi-synthetic double-model setup experiment to calibrate 134 parameters and two well-known distributed hydrological models: the Variable Infiltration Capacity (VIC) model and the Distributed Hydrology Soil Vegetation Model (DHSVM). The results demonstrate statistically that the proposed framework can better mitigate equifinality problem, yields more realistic model parameter estimates, and is computationally more efficient.

stat.AP↗