#139139th of 142 in Agent frameworks and SDKsFor science, health and research, research and papers
paper2code
Multi-agent pipeline that turns ML papers into code repositories
- Stars
- 5k
- License
- Apache-2.0
- Last commit
- Mar 2026
- Language
- Python
4,410 on 21 Apr 20265,019 on 10 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
PaperCoder is the reference implementation of the Paper2Code ICLR 2026 paper. It takes a paper as PDF-derived JSON or LaTeX source and runs three LLM-agent stages (planning, analysis, code generation) to write a code repository. It works with the OpenAI API (o3-mini) or open models served through vLLM, and includes a benchmark dataset and a model-based evaluation script.
Who it is for: ML researchers reproducing papers and people studying paper-to-code generation
Strengths
- Runs against OpenAI API or local vLLM models, default DeepSeek-Coder-V2-Lite-Instruct
- Accepts LaTeX source directly, skipping PDF conversion
- Ships the Paper2Code benchmark on Hugging Face and reference-free and reference-based eval scripts
- Saves planning, analyzing and coding artifacts per paper for inspection
Weaknesses
- PDF input needs a separate s2orc-doc2json and GROBID setup to produce JSON
- Evaluation relies on o3-mini via the OpenAI API, so it is not fully local
- Research code driven by shell scripts; no packaged CLI, release or Docker image
- Scoped to ML papers; output quality varies by model and is not guaranteed
What it needs
- GPU optional
- Needs OpenAI API, vLLM, s2orc-doc2json, GROBID, tiktoken
- Models: OpenAI o3-mini, deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct, vLLM-served open models
Also in Agent frameworks and SDKs
See all 142| Rank | Project | Score |
|---|---|---|
| 1 | ppt-masterAgent skill that turns documents into natively editable PowerPoint decks | 96 out of 100 |
| 2 | openai-agents-pythonPython SDK for multi-agent workflows with handoffs, guardrails and tracing | 92 out of 100 |
| 3 | mastraTypeScript framework for building AI agents, workflows and MCP servers | 92 out of 100 |
| #4 | browser-usePython library that lets an LLM agent control a web browser | 90 out of 100 |
| #5 | copilotkitReact and Angular SDK for adding agent chat and generative UI to apps | 90 out of 100 |