SocraticRepair: Planning-Guided Socratic Tutoring for Program Debugging and Misconception Repair

Key points

• Large language models are increasingly used in computer science education.1
• However, the teaching abilities of even the most powerful LLMs are limited: they tend to reveal the solution prematurely and may lose focus on the learner's current debugging objective.1
• The framework is instantiated in the Python programming domain by fine-tuning Qwen2.5-Coder-7B-Instruct on synthetic multi-turn tutoring dialogues constructed from four preprocessed programming datasets.1
• The experiments show that SocraticRepair achieves higher scores on stepwise scaffolding and question quality while also substantially reducing premature solution disclosure.1

Sources (1)

• [1]SocraticRepair: Planning-Guided Socratic Tutoring for Program Debugging and Misconception Repair
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 10, 10:28 PM

  To address this, we introduce SocraticRepair, a planning-guided Socratic tutoring framework for program debugging and miscon- ception repair that adapts prior planning-guided tutoring work (Li et al., 2026b).

  Large language models are increasingly used in computer science education.

Extractive summary: sentences quoted from the sources.