NiuTrans/LaTeXTrans

What it solves

LaTeXTrans addresses the difficulty of translating academic papers while maintaining the complex formatting, mathematical formulas, and cross-references typical of LaTeX documents. Unlike standard PDF translators that often break layout and equations, this system ensures that the translated output remains a high-fidelity, compilable LaTeX project and PDF.

How it works

The system uses a multi-agent collaboration workflow consisting of six specialized agents:

1. Parser: Handles the initial processing of the LaTeX source.
2. Terminology Extractor: Ensures consistent translation of technical terms.
3. Translator: Performs the actual language translation using an LLM.
4. Validator: Checks the translation for errors.
5. Summarizer: Assists in the overall translation process.
6. Generator: Produces the final translated LaTeX code and compiles it into a PDF.

Users can provide an arXiv ID, a local compressed source package, or a local directory, and the system automatically downloads the source, translates it, and compiles the final PDF.

Who it’s for

Researchers and students who need to read arXiv papers in their native language without losing the original paper's structural integrity and mathematical precision.

Highlights

• End-to-end workflow: Automatically downloads LaTeX source from arXiv via paper ID and generates a translated PDF.
• Layout preservation: Maintains the integrity of formulas, layout, and cross-references.
• Terminology consistency: Uses a dedicated agent to ensure technical terms are translated consistently.
• Flexible input: Supports arXiv IDs, local project files, and batch translation.
• GUI available: Includes a Streamlit-based browser interface for monitoring progress and configuration.