This repository curates research, tools, and resources on Large Language Models (LLMs) applied to scientific discovery, presented as a survey titled "From Automation to Autonomy." It targets researchers, developers, and enthusiasts, offering a structured overview of LLM capabilities in science, from augmenting human tasks to autonomous research. The primary benefit is a comprehensive, categorized compendium of the rapidly advancing field, facilitating informed adoption decisions.

How It Works

The project structures LLM involvement in scientific discovery into a three-level autonomy framework: Level 1 (LLM as Tool) where LLMs assist researchers with specific tasks like literature review or data analysis; Level 2 (LLM as Analyst) where LLMs exhibit greater autonomy in complex information processing and insight generation; and Level 3 (LLM as Scientist) where LLM-based systems autonomously conduct major research stages, potentially culminating in draft papers. This taxonomy provides a clear progression of LLM capabilities and applications within the scientific method.

Quick Start & Requirements

This repository serves as a curated list and survey, not a runnable software project. It provides links to research papers and resources. The primary "resource" is the survey paper itself, available at https://arxiv.org/abs/2505.13259. No installation or specific technical prerequisites are required to access the curated information.

Highlighted Details

• Comprehensive taxonomy categorizing LLM applications in scientific discovery into Tool, Analyst, and Scientist roles.
• Extensive lists of pioneering research papers and tools for each category, covering literature review, hypothesis generation, experiment planning, data analysis, and autonomous research.
• Focus on the evolving autonomy of LLMs, from augmenting human researchers to acting as independent scientific agents.

Maintenance & Community

Contributions of relevant papers, tools, or resources are welcomed via pull requests, following the repository's contribution guidelines. Specific community channels like Discord or Slack are not mentioned.

Licensing & Compatibility

The repository itself does not specify a software license. The survey paper is available via arXiv, typically under terms that permit non-commercial access and sharing. Compatibility for commercial use or integration into closed-source projects would depend on the licenses of the individual research papers and tools listed.

Limitations & Caveats

As a curated survey, the repository reflects the state of research at the time of its compilation and may not include the very latest developments. The rapid evolution of LLMs means that the autonomy levels and capabilities described are subject to ongoing advancement. The repository does not provide executable code or a platform for direct LLM experimentation.