@inproceedings{huang-etal-2025-gumbel,
    title = "{G}umbel Reranking: Differentiable End-to-End Reranker Optimization",
    author = "Huang, Siyuan  and
      Ma, Zhiyuan  and
      Du, Jintao  and
      Meng, Changhua  and
      Wang, Weiqiang  and
      Leng, Jingwen  and
      Guo, Minyi  and
      Lin, Zhouhan",
    editor = "Che, Wanxiang  and
      Nabende, Joyce  and
      Shutova, Ekaterina  and
      Pilehvar, Mohammad Taher",
    booktitle = "Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2025",
    address = "Vienna, Austria",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingestion-acl-25/2025.acl-long.354/",
    pages = "7142--7161",
    ISBN = "979-8-89176-251-0",
    abstract = "RAG systems rely on rerankers to identify relevant documents. However, fine-tuning these models remains challenging due to the scarcity of annotated query-document pairs. Existing distillation-based approaches suffer from training-inference misalignment and fail to capture interdependencies among candidate documents. To overcome these limitations, we reframe the reranking process as an attention-mask problem and propose Gumbel Reranking, an end-to-end training framework for rerankers aimed at minimizing the training-inference gap. In our approach, reranker optimization is reformulated as learning a stochastic, document-wise Top-$k$ attention mask using the Gumbel Trick and Relaxed Top-$k$ Sampling. This formulation enables end-to-end optimization by minimizing the overall language loss. Experiments across various settings consistently demonstrate performance gains, including a 10.4{\%} improvement in recall on HotpotQA for distinguishing indirectly relevant documents."
}