facebookresearch

facebookresearch / autoform-bot

Public

Autoform Bot

19
2
89% credibility
Found May 29, 2026 at 23 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

This project is an AI-powered pipeline from Meta Research that transforms mathematical textbooks written in LaTeX into formally verified proofs using the Lean 4 proof assistant. Multiple AI agents work together: some extract theorems from the book, others write proofs, reviewers check correctness, and a coordinator manages the entire workflow. The system handles failures gracefully by automatically creating fix tasks and alerting humans only when truly stuck. A visual dashboard lets you monitor progress. Think of it as a robot research assistant that reads math textbooks and produces ironclad proofs that mathematicians can trust completely.

How It Works

1
📚 You gather your math textbook

You place your LaTeX or Markdown math book into a folder and the system prepares to read it.

2
🎯 You choose which theorems to formalize

The system extracts all the theorems, definitions, and propositions from your book and lets you confirm which ones to prove.

3
🤖 Your AI workers start proving

Multiple AI assistants work simultaneously, each attempting to prove different statements from your book using the Lean proof language.

4
Two paths to handle failures
🔍
Reviewers check every proof

AI reviewers verify each proof matches the book's statement exactly and follows correct mathematical reasoning.

📝
Fix tasks get created automatically

When something fails, the system breaks down the problem into smaller tasks and tries again.

5
📊 You watch progress on the dashboard

A web dashboard shows you which theorems are proved, which are still pending, and how much work remains.

6
✅ The system verifies correctness

Every successful proof is automatically checked against the mathematical library to ensure it uses only valid reasoning.

🎉 Your book is now formally verified

Your entire textbook has been transformed into a collection of computer-verified mathematical proofs that will never contain errors.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 23 to 19 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 autoform-bot?

Autoform-bot is a multi-agent pipeline for translating LaTeX mathematics into verified Lean 4 proofs. Put a math textbook in, and a swarm of AI agents - workers, reviewers, and orchestrators - collaborates to produce compilable, verified formalizations using Mathlib. The system extracts theorem statements from LaTeX sources, dispatches parallel proof attempts across compute nodes (including SLURM clusters), runs compilation checks, and judges quality through rubric-based grading. A web dashboard lets you inspect traces, monitor progress, and track which theorems have been formalized. It is written in Python and plugs into Anthropic Claude, OpenAI, and Gemini models via their APIs.

Why is it gaining traction?

The hook here is verifiable mathematical knowledge at scale. Math libraries like Mathlib are enormous collaborative efforts where proof verification catches subtle errors humans miss. Automating even part of that process is a holy grail for formal methods research. The tool targets a real bottleneck: formalizing textbooks is tedious, slow, and demands expertise in both the math and the proof assistant. This pipeline abstracts away the distributed compute, retry logic, and evaluation plumbing so researchers can focus on feeding it new source material.

Who should use this?

This is squarely for formal methods researchers, mathematicians working with Lean 4, and contributors to projects like Mathlib who want to bootstrap new formalizations from existing LaTeX sources. If you are building a math textbook or course and want to generate machine-verified proofs alongside it, this is what you reach for. If you want to autoformalize a PDF into a proof assistant without deep Lean expertise, this pipeline handles the orchestration so you do not have to wire it yourself. General developers looking for form automation will want a different project entirely.

Verdict

The credibility score of 0.9% and 19 stars signal that this is a bleeding-edge research artifact, not a production tool. Documentation and test coverage exist, but the project is early, the setup involves compiling Lean toolchains, and the license restricts commercial use. If you are researching mathematical formalization or contributing to Mathlib, it is worth an afternoon to spin up the pipeline and see what it produces on a small example. For everyone else, watch this space or look for higher-maturity alternatives when they emerge.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.