We present ReSuMe, a general framework for mutual enhancement of dense retrieval systems and document summarizers through reinforcement learning. The framework jointly optimizes a language model for generating retrieval-oriented summaries and adapts the retrieval model to these summaries through alternating fine-tuning phases. We employ Group Relative Policy Optimization (GRPO) to fine-tune the language model based on retrieval relevance rather than linguistic quality alone, while the retrieval model is iteratively updated using contrastive learning on the generated summaries. This co-optimization process addresses the fundamental distribution shift problem that arises when retrieval models trained on full documents must operate on synthetic summaries during inference. By progressively reducing this distribution gap, our framework yields two key benefits: improved retrieval performance and a high-quality document summarizer optimized for retrieval tasks. We demonstrate our framework using Contriever on the MS-MARCO dataset, achieving consistent improvements of 13.2% in MRR@10 and 6.7% in Recall@100 over the baseline. The framework is model-agnostic and can be applied to enhance any dense retrieval system while simultaneously producing an effective document summarization model.

Siva Rajesh Kasa, Karan Gupta, Sumegh Roychowdhury, Ashutosh Kumar, Yaswanth Biruduraju, Santhosh Kumar Kasa, Pattisapu Nikhil Priyatam, Arindam Bhattacharya, Shailendra Agarwal, Vijay Huddar. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

Semantic matching is an important component of a product search pipeline. Its goal is to capture the semantic intent of the search query as opposed to the syntactic matching performed by a lexical matching system. A semantic matching model captures relationships like synonyms, and also captures common behavioral patterns to retrieve relevant results by generalizing from purchase data. They however suffer from lack of availability of informative negative examples for model training. Various methods have been proposed in the past to address this issue based upon hard-negative mining and contrastive learning.

Binary classification based methods are commonly used for designing predictive models in healthcare. A common problem in many healthcare datasets is that of imbalance, where there are far more observations in one class than the other during training. In such conditions, most classifiers do not have good predictive accuracy with respect to the under-represented class. We design a new similarity-based classifier to learn from imbalanced datasets, wherein input features are transformed using similarity with respect to a chosen subset of training points. We empirically demonstrate the superiority of our algorithm over state-of-the-art methods for imbalanced data classification in real and synthetic datasets. We also illustrate the application of our classifier in predicting Acute Respiratory Failure (ARF), a critical complication in Intensive Care Units (ICU), using semi-structured text contained in nursing notes recorded during a patient's ICU stay. Our experiments, on more than 800 patient records show that using our new classifier to learn from text- based features can effectively be used to predict ARF and, potentially, other complications in ICUs.