Language Models as Science Tutors

TL;DR

This paper introduces TutorEval and TutorChat, benchmarks and datasets designed to evaluate and improve language models' ability to assist with scientific education and problem-solving involving long texts and multi-disciplinary knowledge.

Contribution

The paper presents new benchmarks and datasets tailored for scientific education tasks, and demonstrates fine-tuned models that outperform existing approaches on these benchmarks.

Findings

Fine-tuned models excel at TutorEval and standard math benchmarks.

Existing dialogue fine-tuning methods perform poorly on scientific long-context tasks.

Open-source datasets and models are released for community use.

Abstract

NLP has recently made exciting progress toward training language models (LMs) with strong scientific problem-solving skills. However, model development has not focused on real-life use-cases of LMs for science, including applications in education that require processing long scientific documents. To address this, we introduce TutorEval and TutorChat. TutorEval is a diverse question-answering benchmark consisting of questions about long chapters from STEM textbooks, written by experts. TutorEval helps measure real-life usability of LMs as scientific assistants, and it is the first benchmark combining long contexts, free-form generation, and multi-disciplinary scientific knowledge. Moreover, we show that fine-tuning base models with existing dialogue datasets leads to poor performance on TutorEval. Therefore, we create TutorChat, a dataset of 80,000 long synthetic dialogues about textbooks. We use TutorChat to fine-tune Llemma models with 7B and 34B parameters. These LM tutors specialized in math have a 32K-token context window, and they excel at TutorEval while performing strongly on GSM8K and MATH. Our datasets build on open-source materials, and we release our models, data, and evaluations.