AWESOME-OCR-LLM: Track OCR Research in the LLM Era

Overview

AWESOME-OCR-LLM is a research-curation resource, not an OCR engine or software library. It organizes papers and related code or model links about text-rich images and documents in the large language model era, making it easier to follow a fast-moving field without treating general document AI or classical OCR as its primary scope.

It is useful when you need a starting point for literature review or want to monitor new work across parsing, document understanding, visual text generation, specialized models, and benchmarks. The repository also flags a field-wide caveat: reported evaluation results often use inconsistent protocols, so scores across papers may not be directly comparable.

Key Features

• Curates OCR and multimodal research focused on text-rich visual content.
• Groups work into document parsing, document understanding, visual text generation, specialized models, and benchmarks and evaluation.
• Provides a daily papers section alongside categorized research lists.
• Includes paper titles, dates, and affiliations in its listings.
• Links to available code, model, or dataset resources where listed.
• Summarizes emerging research directions, including OCR-free understanding, document agents, and evaluation practices.
• States scope boundaries to distinguish this resource from general document-AI, broad MLLM, and classical OCR lists.

Use Cases

• Researchers can use it to find relevant papers when starting a literature review on multimodal OCR or document AI.
• Engineers evaluating document-processing approaches can scan recent methods and follow available project or model links.
• Benchmark designers can use its evaluation coverage to identify work in an area where protocols remain inconsistent.
• Students and newcomers can use its topic categories and trend summary to build an overview of OCR research in the LLM era.

Project Facts

• License: MIT
• Stars: 725
• Forks: 54
• Archived: no

Getting Started

Browse the repository README for the latest curated papers and links. The README points to full categorized lists for additional entries.

Considerations

• This is a manually curated reading list, not a runnable OCR tool, API, or benchmark implementation.
• Research listings and links can change as the repository is updated; check the linked papers and resources for authoritative details.
• The README notes that many reported results rely on self-reported numbers, private test sets, or inconsistent protocols, limiting rigorous cross-paper comparison.
• The input does not specify a programming language or installation process.