G-CARL - Grounded Checklist-Aligned Reward Learning for Patient-Oriented Medical Report Interpretation

Sep 15, 2026

Uploaded by Sage

Personalized interpretation of medical reports is increasingly necessary for patients seeking clear, actionable understanding. Existing medical vision-language methods fall short because they do not simultaneously enforce evidence-grounded medical factuality and context-dependent patient communication. To close this gap, G-CARL introduces Patient-oriented Medical Report Interpretation (PMRI), an open-ended multimodal generation task driven by user queries and dialogue history. The grounded, checklist-aligned reinforcement learning framework uses multi-source retrieval for atomic claim verification and context-aware weighted checklists for coverage and expression quality, supporting patient-safe, accessible explanations. Experiments on MMedReport with clinician evaluation show consistent improvements in overall quality, claim precision, checklist recall, and clinician preference.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages9

Reading Time23 min

AccessFree to download

Introduction

Patient-oriented Medical Report Interpretation (PMRI)

Why supervised fine-tuning is limited

Reinforcement learning reward design challenges

Proposed solution: G-CARL

G-CARL: Grounded Checklist-Aligned Reward Learning for Patient-Oriented Medical Report Interpretation Shiao Xie 1∗, Siyu Chen 1∗, Jianwei Lv 1 , Bo Yuan 1 , Yujin Wang 1†, Xiandong Li 1† 1Baidu Inc., China arXiv : 2608 .2033 1v 1 [cs .CL] 20 Aug 2026 Abstract Personalized interpretation of medical reports has emerged as an increasingly important need among patients. Addressing this need requires both evidence-grounded medical factuality and context-dependent patient communication, yet existing medical vision-language tasks do not adequately capture these dual requirements. To bridge this gap, we introduce Patient-oriented Medical Report Interpretation (PMRI), a novel open-ended multimodal generation task that requires models to explain medical reports in accurate and accessible language based on a user’s query and dialogue history. These two objectives differ fundamentally in their verifiability, yet remain tightly coupled, making them difficult to optimize jointly under conventional supervised fine-tuning and holistic reinforcement learning paradigms. To address this challenge, we propose G-CARL, a grounded, checklist-aligned reinforcement learning framework that combines multi-source retrieval for atomic claim verification with context-aware, instance-specific weighted checklists for response coverage, providing structured supervision for factuality, user-demand satisfaction, and expression quality without constraining response diversity. We further construct MMedReport, a realworld PMRI benchmark, along with a clinician-designed three-dimensional evaluation protocol. Extensive experiments demonstrate that G-CARL consistently outperforms existing post-training baselines in overall quality, claim-level precision, and checklist recall. Pairwise preference evaluation by clinicians further confirms that G-CARL produces interpretations that are more accurate and better aligned with patient needs. 1 Introduction Written within a professional medical context, medical reports commonly present abnormal values and descriptive findings without explaining their implications for non-expert readers. This gap becomes particularly salient in online healthcare scenarios, where patients upload one or more report images and ask open-ended questions shaped by their personal concerns. To bridge this gap, we introduce Patient-oriented Medical Report Interpretation (PMRI), a novel open-ended multimodal generation task. Beyond simply recognizing report findings, PMRI must align professional medical knowledge with patient-centered communi- ∗These authors contributed equally. †Corresponding authors: Yujin Wang and Xiandong Li. Task A : Medical VQA Task B : Medical Report Generation | Task C : Patient-Oriented Medical Report Interpretation | | | | --- | --- | --- | | | Doctor-patient dialogue history

User query | | Figure 1: Comparison of PMRI (Task C) with medical VQA (Task A) and conventional report generation (Task B). Unlike the short, deterministic outputs of Tasks A and B, PMRI produces long-form, patient-facing explanations grounded in multimodal reports and extended patient–doctor dialogue. cation by transforming report evidence into accessible explanations, addressing user-specific concerns, and offering clinically cautious guidance. As illustrated in Fig. 1, PMRI differs from conventional medical visual question answering (VQA) (Lu et al. 2025; Jiang et al. 2025a) and report generation (Lin et al. 2026; Liet al. 2025) in two important ways. First, it is an evidencebounded generation task whose responses must be grounded not only in the reports but also in reliable clinical knowledge and medically coherent diagnostic reasoning, particularly when explaining abnormalities or suggesting follow-up actions. This requirement is essential for patient safety because unsupported interpretations and inappropriate recommendations may mislead patients or delay necessary care. Second, PMRI is a patient-facing communicat

FAQs

What problem does G-CARL target in medical report interpretation?

G-CARL targets the need to generate patient-facing explanations that are both evidence-grounded in the reports and appropriately aligned with patient communication needs, which existing tasks and models do not jointly capture.

How does PMRI differ from medical VQA and conventional report generation?

PMRI is an open-ended multimodal generation task producing long-form, patient-facing interpretations grounded in report evidence and extended patient–doctor dialogue history, unlike short deterministic outputs in VQA or conventional report generation.

What reinforcement learning approach does G-CARL use?

G-CARL uses a grounded, checklist-aligned reinforcement learning framework that verifies atomic claims via multi-source retrieval and supervises response coverage with context-aware, instance-specific weighted checklists to improve factuality, demand satisfaction, and expression quality.

Sep 15, 2026

Uploaded by Sage