Auto-generating question-answering datasets with domain-specific knowledge for language models in scientific tasks

2025/01/01 by Zongqian Li, Jacqueline M. Cole · 1 voice · 3 citations
Computer Science · #Topic Modeling #Natural Language Processing Techniques

paper · pdf · doi:10.1039/d4dd00307a

openalex publication_date 2025/01/01 · openalex created_date 2025/02/25 · openalex updated_date 2026/07/31

Abstract

Algorithms use existing high-quality materials databases to produce a large question-answering dataset whose domain knowledge is sufficient to fine-tune a small language model with high performance.

Citations

• LLM-Prop: Predicting Physical And Electronic Properties Of Crystalline Solids From Their Text Descriptions

Cited by

• A Survey on Prompt Tuning
• MatTools: Benchmarking Large Language Models for Materials Science Tools
• PT-MoE: An Efficient Finetuning Framework for Integrating Mixture-of-Experts into Prompt Tuning

Discussions

• In our next featured article Li and Cole present a pipeline for creating domain-specific language models for scientific tasks without extensive pre-training, by generating a question-answering dataset [bsky, 1 points, 0 comments]

Related

• SciQAG: A Framework for Auto-Generated Science Question Answering Dataset with Fine-grained Evaluation
• Automatic Dataset Generation for Knowledge Intensive Question Answering Tasks
• Prompting-based Synthetic Data Generation for Few-Shot Question Answering
• K-AID: Enhancing Pre-trained Language Models with Domain Knowledge for Question Answering
• A Comparative and Experimental Study on Automatic Question Answering Systems and its Robustness against Word Jumbling