One‑Click LaTeX Template Migration for AI Conference Papers with Auto‑Resubmit

Supported Conferences

ACL family: ACL, EMNLP

NeurIPS family: NeurIPS, NIPS

ICML family: ICML

ICLR family: ICLR

CVPR family: CVPR, ICCV

AAAI family: AAAI

Workflow

Read the source paper zip (the original LaTeX project).

Read the target template zip (the official conference template).

Identify the main entry file (e.g., main.tex).

Extract content: title, abstract, body, references, appendix, and macro definitions.

Rebuild the paper by assembling the extracted parts according to the target template’s structure.

Copy resources such as figures, .bib files, and local style files.

Compile with tectonic and output a new zip package.

Installation and Usage

Environment Requirements

Python 3.10+

pip

tectonic

(for PDF compilation)

Installation

pip install auto-resubmit
# or
git clone https://github.com/LilanOvO/Auto-Resubmit
cd Auto-Resubmit
pip install -e .

Usage

auto-resubmit --source source_paper.zip --target target_template.zip --output output/

Technical Principle

Auto‑Resubmit uses a rule‑based approach consisting of two phases: structured extraction and template reconstruction .

Extraction phase: parses LaTeX source files, identifies the document hierarchy (title, abstract, body, references, appendix) and extracts custom macros and environment definitions.

Reconstruction phase: re‑assembles the extracted elements according to the target template, handling differences such as single‑column versus double‑column layouts.

The implementation relies on the Python standard library; the sole external dependency is tectonic for PDF compilation.

Current Limitations

Highly customized front‑matter or title‑page logic may require manual fixes.

Non‑standard ways of injecting abstract, appendix, or references can break the conversion.

Custom build scripts or external resource dependencies are not handled automatically.

Use Cases

Scenario 1 – Resubmitting to a top‑tier conference: a paper rejected by ACL can be quickly reformatted for EMNLP without manual template adjustments.

Scenario 2 – Simultaneous submissions: when submitting the same work to both ICML and NeurIPS, the tool eliminates duplicated formatting effort.

Scenario 3 – Version management: different camera‑ready versions for multiple conferences can be generated instantly.

GitHub: https://github.com/LilanOvO/Auto-Resubmit