BJTU at BEA 2025 Shared Task: Task-Aware Prompt Tuning and Data Augmentation for Evaluating AI Math Tutors
Yuming Fan, Chuangchuang Tan, Wenyu Song · 2025
We describe the BJTU submission to the BEA 2025 Shared Task on Evaluating the Pedagogical Ability of AI Tutors, which focuses on assessing AI-generated math tutoring responses across four dimensions: Mistake Identification, Mistake Location, Guidance, and Actionability.Our approach leverages a large language model (LLM) with task-specific prompt tuning and data augmentation techniques, including dialogue shuffling and class balancing.The system achieves strong results across all tracks, ranking first in Mistake Identification and performing competitively in the others.Our findings underscore the potential of prompt-based LLMs for pedagogically-aware response evaluation and offer insights into the design of AI tutors with improved educational feedback.