AWESOME-OCR-LLM is a continuously updated repository for tracking recent progress in OCR and visual text understanding in the era of large language models. It focuses on work from roughly the past five years, covering document parsing and understanding, visual text generation, specialized models, benchmarks, and emerging research directions.
The repository mainly collects methods that use large vision-language or multimodal models for text-rich images and documents. It is intentionally focused on OCR-related research rather than serving as a general Document AI or MLLM collection, and classical OCR methods are included only when they are closely connected to this newer line of work.
We also keep track of how recent systems are evaluated. Since many models are released as technical reports and use different datasets, private test sets, or evaluation protocols, reported numbers are often not directly comparable across papers. The repository therefore records results together with their evaluation settings whenever possible, while also highlighting the need for more unified, reproducible, and contamination-resistant benchmarks.
The repository currently includes:
🎉 News
📖 Contents
🔭 Daily Papers
🔍 Emerging Trends
📄 Document Parsing
📄 Document Understanding
📄 Visual Text Generation
📄 Specialized Models
📄 Benchmarks and Evaluation
The repo is updated continuously as new papers, models, and benchmarks appear. Contributions are welcome. For pull requests, please add one entry per model and include the venue or release date, affiliation, and a link to the code or model when available.