facebookresearch

facebookresearch / atlas-lean

Public

ATLAS Autoformalized Textbook Library At Scale

47
5
100% credibility
Found May 29, 2026 at 110 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Lean
AI Summary

ATLAS is a large library of mathematics where undergraduate and graduate textbooks have been automatically translated into machine-verifiable formal code. It covers 26 textbooks across analysis, algebra, geometry, topology, combinatorics, probability, and statistics, with over 42,000 proven theorems. The library is designed to help mathematicians, students, and enthusiasts explore and learn from rigorously verified mathematical results, with a web-based visualizer for easy browsing.

How It Works

1
📚 Discover a Library of Math

You find a massive collection of math textbooks that have been translated into a computer-verifiable format, covering everything from basic algebra to advanced topology.

2
🔍 Browse the Topics

You explore the library organized by subject area—algebra, geometry, probability, and more—to find the math you're interested in.

3
✨ See Proofs Come Alive

You discover that each theorem comes with a machine-checkable proof, so you can trust every statement is mathematically correct.

4
🔗 Follow Dependencies

You trace how one theorem builds on another, seeing the logical connections between results across different areas of math.

5
🌐 Use the Visual Explorer

You visit the web tool that lets you browse theorems, compare informal textbook statements with their formal versions, and extract the code you need.

🎓 Learn with Confidence

You study math knowing every proof has been verified, building your understanding on a foundation of absolute certainty.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 110 to 47 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is atlas-lean?

ATLAS is a Lean 4 library of textbook mathematics autoformalized by LLMs -- informal statements and proofs translated into formal Lean code. It covers 26 textbooks spanning analysis, algebra, geometry, topology, combinatorics, probability, statistics, PDEs, and theoretical computer science. The goal is reusable formal building blocks for human- and machine-driven formalization.

Why is it gaining traction?

The scale is the hook -- 46,203 declarations across real textbooks, with 92.7% of them proved. This demonstrates that LLM autoformalization can handle genuine mathematical content, not just toy examples. The visualizer lets you browse and compare informal statements against their Lean formalizations, making the corpus explorable. Built on Mathlib4 conventions, so it slots into the broader Lean ecosystem.

Who should use this?

Researchers building AI systems for mathematical reasoning will find this valuable as a benchmark and training corpus. Mathematicians exploring formalization might use it as a starting point for theorems already covered. Lean developers working in these specific domains may find reusable material here. This is not production tooling -- it's a research dataset.

Verdict

Experimental. With 47 stars and a 1.0% credibility score, ATLAS is a promising early experiment, not polished software. The math is real and the scale is impressive, but the code is machine-generated and still needs curation. Treat it as a research artifact worth watching, not a dependency to rely on.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.