DocAILab develops open-source research systems and benchmarks for Document AI, Retrieval-Augmented Generation, and intelligent document processing.
Our projects focus on making document-centered AI systems easier to evaluate, reproduce, and deploy. Current work includes modular RAG benchmarking, privacy-preserving federated embedding learning, document fingerprinting and similarity search, and end-to-end intelligent document processing pipelines.
- XRAG: A benchmarking framework for examining core modules in advanced Retrieval-Augmented Generation systems.
- FedE4RAG: Privacy-preserving federated embedding learning for localized RAG retrievers.
- Document-Fingerprints: H3D code and processed datasets for document fingerprinting, similarity search, and fine-grained deduplication.
- IDP-system: An intelligent document processing system.
- Retrieval-Augmented Generation evaluation
- Document representation and fingerprinting
- Privacy-preserving retrieval and federated learning
- Intelligent document processing
- Reproducible benchmarks and open research tools
- Try XRAG if you are evaluating RAG pipelines.
- Try Document-Fingerprints if you are studying document similarity, deduplication, or fingerprint-based retrieval.
- Try FedE4RAG if you are interested in localized RAG under privacy constraints.
- Add runnable quick-start examples for each public repository.
- Add citation, license, and dataset notes to all research repositories.
- Add benchmark tables and reproducibility checklists for released papers.
- Add issue templates for bug reports, feature requests, and paper reproduction questions.
- Pin the most active repositories: XRAG, FedE4RAG, Document-Fingerprints, and IDP-system.
Contributions, issues, and research discussions are welcome.
