Scaling medical imaging report generation with multimodal reinforcement learning

TL;DR

UniRG introduces a reinforcement-learning framework for medical imaging report generation that directly optimizes clinically relevant rewards, yielding robust generalization and factual alignment over supervised fine-tuning. Trained as UniRG-CXR on multi-institution chest X-ray datasets using a two-stage SFT+RL pipeline, it achieves state-of-the-art results on the ReXrank leaderboard across multiple datasets and report settings. The evaluation spans report-level metrics, disease-level diagnostic accuracy, longitudinal consistency, zero-shot generalization to unseen datasets, and demographic fairness, demonstrating universal performance benefits across institutions, metrics, diagnostic levels, temporal contexts, and patient populations. This work positions UniRG-CXR as a universal foundation for radiology report generation with the potential to generalize to real-world clinical workflows and multimodal data integration.

Abstract

Frontier models have demonstrated remarkable capabilities in understanding and reasoning with natural-language text, but they still exhibit major competency gaps in multimodal understanding and reasoning especially in high-value verticals such as biomedicine. Medical imaging report generation is a prominent example. Supervised fine-tuning can substantially improve performance, but they are prone to overfitting to superficial boilerplate patterns. In this paper, we introduce Universal Report Generation (UniRG) as a general framework for medical imaging report generation. By leveraging reinforcement learning as a unifying mechanism to directly optimize for evaluation metrics designed for end applications, UniRG can significantly improve upon supervised fine-tuning and attain durable generalization across diverse institutions and clinical practices. We trained UniRG-CXR on publicly available chest X-ray (CXR) data and conducted a thorough evaluation in CXR report generation with rigorous evaluation scenarios. On the authoritative ReXrank benchmark, UniRG-CXR sets new overall SOTA, outperforming prior state of the art by a wide margin.

Figures (4)

• Figure 1: Overview of UniRG-CXR. (a) Training Data: UniRG-CXR is trained on the training splits of MIMIC-CXR mimic-cxr, CheXpert Plus chexpert-plus, ReXGradient-160k rexgradient and IU iuxray covering diverse institutions and patient demographics. (b) Training and Rewards: Taking input from the current image, clinical context (e.g., indication), and optionally prior studies, UniRG-CXR uses GRPO reinforcement learning to optimize composite rewards that combine rule-based, model-based, and LLM-based metrics. (c) Evaluation: We assess UniRG-CXR on held-out test sets from MIMIC-CXR, CheXpert Plus, ReXGradient and additionally assess zero-shot generalization on an external proprietary dataset and IU-Xray (the zero-shot setting excludes IU-Xray training data). Report quality is measured using ReXrank metrics rexrank and an LLM-based clinical-error metric zambrano2025clinically, while diagnostic ability is evaluated via F1-based disease classification from generated reports. (d) ReXrank Results: UniRG-CXR achieves SOTA performance across four datasets and two generation settings (findings only and findings + impression), showing substantial gains over prior state-of-the-art.
• Figure 2: UniRG-CXR achieves state-of-the-art performance, delivering consistent and comprehensive performance gains across metrics. (a) On the ReXrank leaderboard, UniRG-CXR (green) shows robust, universal improvement across all evaluation metrics. (b). Starting from the same SFT checkpoint, RL with our combined reward achieves more balanced gains across metrics and the highest RadCliQ-v1 score compared to RL on single metrics. This ablation study is trained and tested on MIMIC (c). Ablation study on the training dynamics shows RL full (UniRG-CXR) achieves significantly better RadCliQ-v1 score than RL only on BLEU. (d). During training, RL full (UniRG-CXR) shows a steady decrease in clinical errors per report as compared with a fluctuating trajectory without consistent improvement from an ablation run without error awareness (i.e. removing CheXprompt metric optimization). Both (c) and (d) show results on 1024 MIMIC validation set from ablations that are trained on MIMIC. (e). Case studies illustrate that UniRG-CXR can produce error-free reports, unlike MedVersa and MedGemma. (f). UniRG-CXR yields a substantially higher proportion of reports with $\leq 1$ error and fewer with $\geq 4$ errors than prior models.
• Figure 3: UniRG-CXR enhances longitudinal report generation. (a). Comparing UniRG-CXR and its non-longitudinal ablation with prior models on longitudinal report generation, we show UniRG-CXR exhibits the best performance and the longitudinal information is beneficial to the performance. (b). UniRG-CXR achieves the best performance across different longitudinal encounter points ranging from the first encounter to the more complex 5th+ encounters, showcasing its improvements are across the board. In comparison, prior models such as GPT-5, GPT-4o and MedGemma are barely surpassing the copy prior report baseline (grey lines). (c). Compared with prior models which barely improve over the copy prior baseline (dashed line), UniRG-CXR significantly and consistently improves performance across different temporal disease change categories including new development, no change, progression and regression (categorized by GPT-5 on ground truth report). Qualitative examples are shown for each category where UniRG-CXR correctly predicts the temporal change based on the input. All results in this figure are on MIMIC test set with prior information where available.
• Figure 4: Generalization and robustness of UniRG-CXR. (a). We held out two datasets sources (IU-Xray and PD (proprietary data) from the training data and evaluate UniRG-CXR in a zero-shot setting on these datasets . UniRG-CXR consistently outperforms prior models, maintaining substantial performance gains in this challenging setup. (b) and (c) present condition-level F1 scores on MIMIC-CXR and PD and highlight that UniRG-CXR remains the overall top-performing model in condition-level diagnostic accuracy. (d). UniRG-CXR demonstrates stable and robust performance across gender, age, and race subgroups, all of which exceed the performance of the second-best model (the dashed lines).