Paper2Code

Automating code generation from scientific ML papers using LLMs

Paper2Code automates the process of converting machine learning research papers into runnable code — bridging the gap between theory and reproducibility.

Motivation

ML research reproducibility is a known challenge. Paper2Code reduces the barrier by automatically extracting algorithms and experimental setups from papers and generating corresponding implementation code via LLMs.

Approach

  • Parse scientific paper structure (abstract, methods, experiments)
  • Extract algorithmic pseudocode and architecture descriptions
  • Generate runnable Python/PyTorch implementations via LLM prompting
  • Validate against paper-reported results where possible

Tech Stack

Python · LLMs · PDF Parsing · PyTorch · LangChain