ALDbench
收藏资源简介:
ALDbench是由阿贡国家实验室的研究团队创建的一个开放式问题基准数据集,专门用于评估大型语言模型(LLMs)在材料合成领域,特别是原子层沉积(ALD)技术中的表现。该数据集包含70个问题,涵盖从研究生水平到领域专家级别的难度,涉及ALD技术的多个方面,如材料生长、过程细节、一般知识及应用。数据集的创建过程由六位ALD领域的专家共同完成,问题经过精心设计,确保每个问题具有可验证的答案。ALDbench主要用于测试LLMs在材料合成中的知识深度和研究辅助能力,旨在解决材料科学领域中复杂问题的自动化处理问题。
ALDbench is an open-ended question benchmark dataset developed by a research team at Argonne National Laboratory, specifically tailored to evaluate the performance of Large Language Models (LLMs) in the field of materials synthesis, particularly atomic layer deposition (ALD) technology. This dataset contains 70 questions spanning difficulty levels from graduate-level to domain expert-level, covering multiple aspects of ALD technology such as material growth, process details, general knowledge and practical applications. The dataset was co-created by six ALD domain experts, and all questions were meticulously designed to ensure each has a verifiable answer. ALDbench is primarily used to test the depth of domain knowledge and research support capabilities of LLMs in materials synthesis, aiming to address the automated processing of complex problems in materials science.

- 1Benchmarking large language models for materials synthesis: the case of atomic layer deposition阿贡国家实验室 · 2024年



