@inproceedings{abdallah-etal-2026-rankify,
    title = "Rankify: A Comprehensive Python Toolkit for Retrieval, Re-Ranking, and Retrieval-Augmented Generation",
    author = "Abdallah, Abdelrahman  and
      Piryani, Bhawna  and
      Mozafari, Jamshid  and
      Herzinger, Andreas  and
      Holdcroft, Jamie  and
      Jatowt, Adam",
    editor = "Durrett, Greg  and
      Jian, Ping",
    booktitle = "Proceedings of the 64th Annual Meeting of the {A}ssociation for {C}omputational {L}inguistics (Volume 3: System Demonstrations)",
    month = jul,
    year = "2026",
    address = "San Diego, California, United States",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-acl/2026.acl-demo.21/",
    pages = "208--219",
    ISBN = "979-8-89176-392-0",
    abstract = "Building retrieval-augmented generation (RAG) systems often requirescombining separate tools for retrieval, re-ranking, and generation,with incompatible data formats, evaluation pipelines, and deployment workflows.We present , an open-source Python toolkit that unifies these stagesin a single modular framework.[PyPI: {\ensuremath{<}}https://pypi.org/project/rankify/{\ensuremath{>}}]{,}[GitHub: {\ensuremath{<}}https://github.com/DataScienceUIBK/Rankify{\ensuremath{>}}]{,}[Docs: {\ensuremath{<}}https://rankify.readthedocs.io{\ensuremath{>}}]{\%}{,}[Video: {\ensuremath{<}}https://youtu.be/kkLzomrM2ec{\ensuremath{>}}]provides 42 benchmark datasets with pre-retrieved documents andpre-built indices, 15 retrievers (sparse, dense, and reasoning-augmented),and 24 re-ranking models spanning 41 pointwise, pairwise, and listwise variants.It also supports \textbf{6} RAG strategies across four inference backends(Hugging Face, vLLM, LiteLLM, and OpenAI), enabling consistent experimentationfrom local models to hosted APIs.A unified pipeline interface allows users to compose retrieve{--}rerank{--}generateworkflows in a few lines of code, while an agentic assistant (RankifyAgent), aREST server (RankifyServer), and an interactive webplayground support deployment and non-programmatic exploration.Across 200+ configurations on QA and BEIR/TREC benchmarks with six generator LLMs,re-ranking consistently improves downstream performance, yielding gains of5{--}15 points in Exact Match and up to 8.5 points in RAGAS context precisionacross diverse retriever{--}generator combinations."
}