LAILab at ArchEHR-QA 2025: Test-time scaling for evidence selection in grounded question answering from electronic health records
Year: 2025
Type: conference-paper
Abstract: This paper presents our approach to the ArchEHR shared task on generating answers to real-world patient questions grounded in evidence from electronic health records (EHRs).We investigate the zero-shot capabilities of general-purpose, domain-agnostic large language models (LLMs) in two key aspects: identifying essential supporting evidence and producing concise, coherent answers.To this aim, we propose a two-stage pipeline: (1) evidence identification via test-time scaling (TTS) and (2) generating the final answer conditioned on selected evidences from the previous stage.Our approach leverages high-temperature sampling to generate multiple outputs during the evidence selection phase.This TTS-based approach effectively explores more potential evidences which results in significant improvement of the factuality score of the answers.
Authors: Tuan Dung Le, Thanh Duong, Shohreh Haddadan, Behzad Jazayeri, Brandon Manley, Thanh Thieu
Institutions: University of South Florida, Moffitt Cancer Center
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4412889079