TutorMoments: Do AI tutors know when to help and when to hold back?
Hugging Face
Read full postAllen Institute researchers introduced TutorMoments, a framework evaluating if large language models can balance helping students and encouraging independent reasoning in math tutoring. Using real tutoring transcripts, they found LLMs often over-assist, rarely prompting deeper student thinking. Providing explicit instructions improves model behavior but doesn't match human tutors' adaptability.



