SciQAG: A Framework for Auto-Generated Science Question Answering Dataset with Fine-grained Evaluation

TL;DR

SciQAG is a framework that automatically generates a large, high-quality science QA dataset from scientific literature using LLMs, and introduces a benchmark to evaluate LLMs' scientific question-answering capabilities.

Contribution

The paper presents SciQAG, a novel framework for automatic science QA dataset generation and a new benchmark for evaluating LLMs in scientific question answering.

Findings

Fine-tuning LLMs on SciQAG improves performance on scientific QA tasks.

The dataset contains 188,042 QA pairs from 24 scientific domains.

The framework enables large-scale, research-level question generation from scientific literature.

Abstract

We introduce SciQAG, a novel framework for automatically generating high-quality science question-answer pairs from a large corpus of scientific literature based on large language models (LLMs). SciQAG consists of a QA generator and a QA evaluator, which work together to extract diverse and research-level questions and answers from scientific papers. Utilizing this framework, we construct a large-scale, high-quality, open-ended science QA dataset containing 188,042 QA pairs extracted from 22,743 scientific papers across 24 scientific domains. We also introduce SciQAG-24D, a new benchmark task designed to evaluate the science question-answering ability of LLMs. Extensive experiments demonstrate that fine-tuning LLMs on the SciQAG dataset significantly improves their performance on both open-ended question answering and scientific tasks. To foster research and collaboration, we make the datasets, models, and evaluation codes publicly available, contributing to the advancement of science question answering and developing more interpretable and reasoning-capable AI systems.