Paper evaluates SHAP and LLM rationales for teaching quality scoring

Arxiv CS.CL (NLP/language models)

Proposes using SHAP and LLM-generated explanations to interpret automated rubric-based scoring of classroom transcripts. Study compares interpretability methods for complex language performance assessment.

Arxiv CS.CL (NLP/language models)·Ivo BuenoBabette B\"uhlerPhilipp StarkTim F\"uttererUlrich TrautweinDorottya DemszkyHeather HillEnkelejda Kasneci

6 days ago