Sherpa: Teaching LLMs to Teach Adaptively
Sherpa uses multi-turn reinforcement learning so LLM teachers adapt instruction to simulated students' learning outcomes.
Sherpa is a multi-turn reinforcement learning framework that trains LLM teachers by maximizing learning outcomes of simulated student archetypes with different preferences. Teacher models improved student performance across archetypes by an average of 20.5 percentage points. On MathTutorBench, the overall pedagogy score rose from 52.5% to 79.2%. In human studies, the trained teacher was preferred over the base model in 79.6% of pairwise comparisons.
- Sherpa trains teachers against multiple simulated student archetypes.
- Average student performance rose 20.5 percentage points.
- MathTutorBench pedagogy score increased from 52.5% to 79.2%.
- Humans preferred the trained teacher in 79.6% of comparisons.
- Products
- MathTutorBench
- AI models
- Sherpa
From the original article
1 min read at huggingface.co
Large language models (LLMs) have become increasingly capable problem solvers, but being able to solve a problem is not the same as being able to teach it. Existing approaches to training LLMs as teachers rely on demonstrations, preference data, or predefined pedagogical criteria that specify what good teaching looks like.
By Weixian Xu, Yanzhe Zhang, Zora Zhiruo Wang · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.