This repository serves as a comprehensive survey of Scientific Large Language Models (Sci-LLMs), focusing on their applications in biology and chemistry. It aims to consolidate research, datasets, and benchmarks for researchers and practitioners in these specialized AI domains, providing a structured overview of the rapidly evolving field.

How It Works

The survey categorizes Sci-LLMs based on their primary data modalities and application areas: Textual (medical, biology, chemistry), Molecular (property prediction, generation, reaction prediction), Protein, Genomic, and Multimodal (combining different data types). It meticulously lists relevant papers, datasets, and benchmarks within each category, offering a structured landscape of the Sci-LLM ecosystem.

Quick Start & Requirements

This repository is a curated survey and does not have direct installation or execution requirements. It provides links to papers, code, and datasets for further exploration.

Highlighted Details

• Extensive coverage of LLMs across biological and chemical domains, including specialized areas like molecular, protein, and genomic data.
• Detailed listing of relevant datasets and benchmarks for evaluating Sci-LLM performance.
• Categorization of models based on their input modalities (text, molecule, protein, genome) and tasks.
• Regular updates to incorporate the latest research, with the latest version available on arXiv.

Maintenance & Community

The project is maintained by HICAI-ZJU and lists several contributors. Users are encouraged to recommend missing papers via issues or pull requests. Contact information for Xinda Wang is provided.

Licensing & Compatibility

The repository itself is a survey and does not impose licensing restrictions. Individual papers and code linked within the survey will have their own respective licenses.

Limitations & Caveats

As a survey, this repository does not provide executable code or models. Its value is in its comprehensive cataloging of existing research, requiring users to independently access and evaluate the linked resources.