Work

Budget-Aware Routing for Long Clinical Text

Year: 2026

Type: conference-paper

Abstract: , a monotone submodular objective that balances relevance, coverage, and diversity. We compare sentence, section, window, and cluster-based unitization, and introduce a routing heuristic that adapts to the budget regime. Experiments on MIMIC discharge notes, Cochrane abstracts, and L-Eval show that optimal strategies depend on the evaluation setting. Positional heuristics perform best at low budgets in extractive tasks, while diversity-aware methods such as MMR improve LLM generation. Selector choice matters more than unitization, with cluster-based grouping reducing performance and other schemes behaving similarly. ROUGE saturates for LLM summaries, while BERTScore better reflects quality differences. We release our code at https://github.com/stone-technologies/ACL_budget_paper.

Source: Findings of the Association for Computational Linguistics: ACL (Findings of ACL)

DOI: 10.18653/v1/2026.findings-acl.2114

Authors: Khizar Qureshi, Geoffrey Martin, Yifan Peng

Institutions: Massachusetts Institute of Technology, Cornell University, Weill Cornell Medicine

Language: English

Topic: Advanced Text Analysis Techniques

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: hybrid

Funder: U.S. National Library of Medicine

Awards: ClinEX - Clinical Evidence Extraction, Representation, and Appraisal, Closing the loop with an automatic referral population and summarization system, Extraction and summarization of evidence-based medicine