[Paper Review] Automated Extraction of Multicomponent Alloy Data Using Large Language Models for Sustainable Design

Arjun Sundaram, Mohit Chakraborty+3 more|arXiv (Cornell University)|Feb 4, 2026

Machine Learning in Materials Science0 citations

TL;DR

The paper develops an LLM-based two-stage pipeline to extract alloy data from text and tables across HEA literature, building large databases for sustainable materials design and demonstrating application to three domains.

ABSTRACT

The design of sustainable materials requires access to materials performance and sustainability data from literature corpus in an organized, structured and automated manner. Natural language processing approaches, particularly large language models (LLMs), have been explored for materials data extraction from the literature, yet often suffer from limited accuracy or narrow scope. In this work, an LLM-based pipeline is developed to accurately extract alloy-related information from both textual descriptions and tabular data across the literature on high-entropy (or multicomponent) alloys (HEA). Specifically two databases with 37,711 and 148,069 entries respectively are retrieved; one from the literature text, consisting of alloy composition, processing conditions, characterization methods, and reported properties, and other from the literature tables, consisting of property names, values, and units. The pipeline enhances materials-domain sensitivity through prompt engineering and retrieval-augmented generation and achieves F1-scores of 0.83 for textual extraction and 0.88 for tabular extraction, surpassing or matching existing approaches. Application of the pipeline to over 10,000 articles yields the largest publicly available multicomponent alloy database and reveals compositional and processing-property trends. The database is further employed for sustainability-aware materials selection in three application domains, i.e., lightweighting, soft magnetic, and corrosion-resistant, identifying multicomponent alloy candidates with more sustainable production while maintaining or exceeding benchmark performance. The pipeline developed can be easily generalized to other class of materials, and assist in development of comprehensive, accurate and usable databases for sustainable materials design.

Motivation & Objective

• Motivate the need to convert unstructured literature into structured, machine-readable data for sustainable materials design.
• Develop a robust, generalizable data extraction pipeline using LLMs that handles text and tabular data across diverse alloy reporting styles.
• Create two comprehensive databases from text and tables, enabling downstream sustainability-aware materials selection.
• Publicly release the curated database to support broader research in HEAs and beyond.

Proposed method

• Two-stage extraction pipeline: (i) paragraph-level text extraction to capture alloy systems, processing, characterization, and properties; (ii) table-based extraction to capture property values, units, and conditions.
• Query Set 1 (QS1) uses prompt engineering, few-shot demonstrations, and retrieval-augmented generation (RAG) to identify alloy compositions, processing, and properties from abstracts and experimental sections.
• Query Set 2 (QS2) maps table cells to a curated master property vocabulary of 354 items, using a two-pass LLM approach to first identify standardized property names and then extract corresponding values and conditions.
• A master property vocabulary is built from DB1 with normalization and an expanded symbol/name set to enable robust table mapping.
• Evaluation uses an extended confusion-matrix framework to account for omissions, hallucinations, and new entries, highlighting precision/recall/F1 trade-offs.
• The pipeline selects GPT-4o and GPT-4o mini for cost/accuracy balance, and applies RAG with 98 expert-annotated examples embedded in vector databases for few-shot demonstrations.

Figure 1: Overall LLM-based alloy data extraction pipeline developed in this work. Starting from list of article DOIs, different sections of the article were separately extracted in XML format using publisher APIs. This data was selectively passed through two LLM extractor consisting of two query se

Experimental results

Research questions

• RQ1 Can an LLM-based pipeline accurately extract alloy compositions, processing details, and properties from both text and tables across a large HEA literature corpus?
• RQ2 What is the achievable precision, recall, and F1 for text (QS1) and table (QS2) extraction compared to expert benchmarks?
• RQ3 How large and usable are the resulting databases, and can they inform sustainability-aware materials selection in multiple domains?
• RQ4 What are the practical challenges and limitations of LLM-based extraction for multicomponent alloys, and how can they be mitigated?

Key findings

• Two databases were created: 37,711 text-derived alloy records (DB1) and 148,069 table-derived records (DB2) from 10,829 articles.
• QS1 achieved an F1 score of approximately 0.83 on expert-annotated review data.
• QS2 achieved an F1 score of approximately 0.88 on a broad test set and 0.96 on a mechanical-property-focused set.
• Text extraction precision/recall on review data were 0.81 and 0.86, respectively, for QS1; table extraction precision/recall were 0.98 and 0.81, respectively.
• The assembled database enables sustainability-aware selection in lightweight structural, soft magnetic, and corrosion-resistant domains, identifying multicomponent alloys with improved sustainability without sacrificing performance.
• The Alloy Tattvasar platform provides public access to the curated database for reuse by the community.

Figure 2: a) Schematic of the QS1 prompt structure, comprising system and formatting instructions, domain-specific context, RAG-selected few-shot examples, and user query with the target article paragraph, organized in a chain-of-thought framework. b) QS1 data extraction workflow for a single articl