GYPCC2/korean-tutoring-persona-dataset

Korean Tutoring Persona Dataset --- Overview --- This dataset is constructed to evaluate the impact of learner persona modeling in bilingual Korean tutoring systems.

Each instance includes:

• —Learner persona information
• —Task type (grammar, error correction, multi-turn dialogue)
• —Input query
• —Three responses generated under different persona configurations --- Data Structure --- Each data instance is formatted as:

• —persona: learner profile (TOPIK level, native language, etc.)
• —task: task type
• —query: input question
• —response_R1: full persona response
• —response_R2: partial persona response
• —response_R3: no persona response --- Task Types ---
• —Grammar Explanation
• —Error Correction
• —Multi-turn Dialogue --- Purpose --- This dataset is designed for:
• —Evaluating instructional consistency
• —Analyzing difficulty alignment
• —Measuring error-targeted feedback

Citation --- If you use this dataset, please cite:

@article{ title={Evaluating the Impact of Learner Persona Modeling on Instructional Consistency and Adaptive Feedback in Bilingual AI Tutoring Systems}, author={Yang, Cheng}, year={2026} }