Large Language Models and Instructional Moves: A Baseline Study in Educational Discourse
Published:Dec 24, 2025 05:00
•1 min read
•ArXiv NLP
Analysis
This ArXiv NLP paper investigates the baseline performance of Large Language Models (LLMs) in classifying instructional moves within classroom transcripts. The study highlights a critical gap in understanding LLMs' out-of-the-box capabilities in authentic educational settings. The research compares six LLMs using zero-shot, one-shot, and few-shot prompting methods. The findings reveal that while zero-shot performance is moderate, few-shot prompting significantly improves performance, although improvements are not uniform across all instructional moves. The study underscores the potential and limitations of using foundation models in educational contexts, emphasizing the need for careful consideration of performance variability and the trade-off between recall and precision. This research is valuable for educators and developers considering LLMs for educational applications.
Key Takeaways
Reference
“We found that while zero-shot performance was moderate, providing comprehensive examples (few-shot prompting) significantly improved performance for state-of-the-art models...”