modelscope/Awesome-Vibe-Research

What it solves

Researchers at any stage of a project often waste time hunting for the right AI tools, evaluating whether they are reliable, and stitching together disconnected scripts. This project solves that by giving the community a single, continuously updated place to find and share AI-assisted workflows that have been tried in real research. It organizes tools by the phase of research they help with, so you can quickly see what exists for literature review, data preparation, experiment execution, paper writing, and more.

How it works

The repository is a structured table-based catalog. It splits the research lifecycle into 10 stages, from end-to-end automation (stage 0) through direction scanning, literature research, data prep, method design, experiment execution, visualization, writing, reproducibility, and dissemination. For each stage, it lists curated projects with a short description, the type of component (agent, skill, tool, workflow, python package, plugin, platform), a GitHub link, star count, and optional links to demos or papers. The project also includes a best-practices section with concrete tutorials and encourages contributions. Community members can add a row for a tool they have used, or share a reusable workflow they developed themselves, including failure cases and boundaries. Star counts are automatically updated daily by a GitHub Action.

Who it’s for

This is for researchers, scientists, and graduate students who want to use AI to accelerate any part of their research process, and who value tools that are open-source, actively maintained, and have clear documentation. It is also for developers or tool builders who want to see what is missing and contribute their own tested workflows.

Highlights

• Covers the entire research lifecycle in one place, with 10 clearly defined stages.
• Each entry includes a concise description, type (agent, skill, tool, etc.), GitHub link, star count, and often a demo or paper reference.
• Curated with strict inclusion criteria: research-focused, runnable code or reusable process, actively maintained, clear docs, and proven in practice.
• Community-driven: anyone can add a line for a tool they used or share their own worked example, including failure cases.
• Includes a best-practices section with real tutorials, such as using AlphaFold3 for protein structure inference and natural-language literature search.
• Star counts are automatically refreshed daily, so the list stays up to date without manual effort.