Medical
-
Generalist-Specialist Mixture-of-Experts for Rare Pathology Detection in Multimodal Imaging
2026-09-16 14:04:24Johannes Kaiser, Florian Braunmiller, Daniel R\"uckert, Georgios KaissisAbstract
AI models for multimodal medical imaging must balance modality-specific specialization with cross-modal shared representations, a trade-off that pure Mixture-of-Experts (MoE) architectures currently fail to satisfy. Expert-based routing improves in-domain learning but may sacrifice cross-modal signals, which appear particularly important for rare (low-prevalence) pathologies in our experiments. To resolve this, we introduce Generalist-Specialist-MoE (GS-MoE), a two-branch (MoE) architecture that couples a cross-modal generalist model with distinct modality-specific specialists (experts) via domain-constrained feature fusion. On RadImageNet (1.35M images, 165 pathologies, three modalities), GS-MoE recovers detection of six low-prevalence pathologies on which every baseline scores F1 $=$ 0, with per-class gains up to +0.60 F1. It attains this while even slightly exceeding dense and specialist-only MoE aggregate baselines (MCC 0.770), while using ${\sim}53\%$ fewer active parameters at inference than the strongest investigated dense model.
Abstract (translated)
URL
https://arxiv.org/abs/2609.18688
PDF
-
Hyperbolic Graph Representation Learning for Differential Diagnosis on Biomedical Knowledge Graphs
2026-09-16 11:11:46Pietro Miotto, Lucia Mellini, Tommaso Marzi, Cesare Alippi, Elena Casiraghi, Alberto Paccanaro, Giorgio Valentini, Mauricio Soto-GomezAbstract
Biomedical knowledge graphs combine ontology-derived hierarchies with transversal associations among heterogeneous entities such as phenotypes, diseases, genes, proteins, and patients. This hybrid structure raises the question of whether hyperbolic embeddings, which naturally capture tree-like organization, remain useful beyond purely hierarchical graphs. We present a preliminary study of hyperbolic graph representation learning for Mendelian-disease differential diagnosis on a patient-integrated biomedical graph. Experiments on isolated ontology subgraphs show that hyperbolic models achieve strong performance in substantially lower dimensions than Euclidean baselines. We then evaluate the models on a link-prediction task that ranks candidate diseases for each patient. Results suggest that hyperbolic embeddings can exploit biomedical hierarchical structure while supporting diagnostic reasoning over heterogeneous patient-level graphs.
Abstract (translated)
URL
https://arxiv.org/abs/2609.18481
PDF
-
A Non-Linear Neuron Based Detection of Isolated Pixels in Binary and Grayscale Images using Contrast Sensitive Receptive Fields
Abstract
Identifying isolated points is important in image processing applications such as medical imaging, astronomy and quality control management. Other domains, such as cybersecurity, also present challenges that can be framed as image processing problems. One example of particular interest is the identification of anomalous single nodes in spatially organised networks where groups of nodes in different regions share similar feature values. This task can involve both binary and more complex grayscale images. However, existing methods face limitations: template matching is infeasible for grayscale images, while 2nd order derivative based methods are highly sensitive to noise and require user-specified thresholds. To overcome these issues, a novel method is proposed for detecting meaningful single-pixel deviations in images. This approach modifies and extends a neuron model, originally designed for anomaly detection, to operate on spatially diameter limited receptive fields that incorporate excitatory and inhibitory regions. The result is a method that is free from user-specified thresholds and parameters, and can be applied to both binary and grayscale images, providing an effective, robust and efficient solution.
Abstract (translated)
URL
https://arxiv.org/abs/2609.18399
PDF
-
REPAIR: Resolving Long-Tail Confusion in Scientific Retrievers via Fact-Verified Iterative Refinement
Abstract
Precise retrieval of scientific information is fundamentally constrained by long-tailed concepts and high fact-sensitivity of scientific corpora. These challenges often limit the effectiveness of dense retrievers and hallucination-prone LLM augmentation. To address this, we present REPAIR, a self-evolving data augmentation framework for scientific dense retrievers. REPAIR iteratively synthesizes training data to address knowledge gaps by cycling through diagnosis of long-tail concepts, API-guided evidence expansion, and differentiation via hard negative mining. This process effectively grounds retrieval in factual reality to resolve fine-grained distinctions. Extensive experiments demonstrate that REPAIR significantly outperforms 19 strong baselines on nine materials science and biomedical benchmarks. Our work highlights that diagnosing and factually augmenting data to long-tail deficits is essential for robust scientific retrieval.
Abstract (translated)
URL
https://arxiv.org/abs/2609.18262
PDF
-
Evolving Error States: Failure-Aware Progressive Repair for Ultrasound Lesion Segmentation
2026-09-16 07:35:19Ziliang Wang, XuJiang Tang, Lu Yuting, Weixin Xu, Yongqiang Zhao, Ying Fu, Kehua GuoAbstract
Reliability under sparse and heterogeneous failures remains a fundamental challenge for medical image segmentation. High average accuracy can conceal a small set of structurally distinct and clinically consequential errors. Existing post-hoc correction methods alleviate this problem, but typically estimate false-positive and false-negative corrections from the same fixed prediction. This ignores the dynamic evolution of error states and limits the correction of complex cases. Inspired by iterative error feedback in structured prediction, we propose Failure-Aware Progressive Repair (FAPR). FAPR represents the current segmentation mask as a dynamic failure state and models each repair operation as a state-transition operator. Each accepted correction forms a new prediction state for subsequent error diagnosis and repair, enabling later operations to adapt to preceding changes. Conditional routing selectively activates necessary state transitions, while failure replay exposes the model to rare error states. By keeping the base segmentor frozen, FAPR preserves its established segmentation capability while improving difficult cases. Across three public ultrasound lesion segmentation benchmarks, FAPR improves mean DSC by 1.52%. On the very-hard subsets of BUSI and TN3K, the average gain reaches 13.77%.
Abstract (translated)
URL
https://arxiv.org/abs/2609.18256
PDF
-
A Lightweight CNN Integrated Compact Convolutional Transformer for Multi-Scale Feature Learning and reducing computational complexity for breast cancer mammography image detection and classification
2026-09-16 06:42:38Md Taimur Ahad (Department of Management North South University, Dhaka, Bangladesh), Ainuddin Ahmed (Department of Management North South University, Dhaka, Bangladesh)Abstract
Over the years, Convolutional Neural Networks (CNNs) have demonstrated strong capability in cancer detection and classification using medical images. However, CNN-based models often struggle to capture long-range contextual dependencies. In such scenarios, integrating Compact Convolutional Transformer (CCT) architectures after the CCT layer allows CNN-extracted features to reshape into compact patch tokens using a CCT tokenizer, followed by the addition of positional embeddings to preserve spatial structure. Using 5-fold cross-validation, the model was tested on 3 sets of breast cancer mammography. With only 250,435 parameters, the model achieved 99%-100% accuracy across 3 datasets, indicating robust generalization. Explainable AI (XAI) was integrated into the model to explain the breast cancer classification process to enhance clinical trust. The results indicate that the proposed framework is suitable for computer-aided diagnosis systems, particularly in resource-constrained clinical environments. The novelty of the proposed CNN-integrated CCT overcomes the limitation of CNN's gradient degradation in the last layers by integrating convolutional tokenization with transformer-based learning. Lighter than ViT, which is effective in capturing long-range dependencies, the model has also proven efficient in breast cancer classification by capturing long-range dependencies among breast tissue regions.
Abstract (translated)
URL
https://arxiv.org/abs/2609.18212
PDF
-
SNOMED CT Concept Recommendation from Masked Clinical Context
Abstract
Standardizing clinical language to SNOMED CT supports interoperability, analytics, and reusable phenotyping, but concept recommendation remains difficult when relevant concepts are rare or absent from training data. We present a masked-concept recommendation benchmark using the SNOMED CT Entity Linking Challenge v1.2.1 data derived from MIMIC-IV-Note. The dataset contains 75,491 annotations across 272 discharge summaries, with 204 notes used for training and 68 for historical testing. For each unique note-concept pair, the target mention is masked from a local clinical context and the system ranks SNOMED CT concepts observed during training. We compare a popularity baseline, sparse TF-IDF concept prototypes, dense latent semantic analysis embeddings, sparse-dense fusion, retrieved-note evidence, and a retrieval-augmented hybrid. Sparse TF-IDF performs best, achieving Recall@1 of 14.81%, Recall@10 of 33.43%, MRR of 0.2114, and nDCG@10 of 0.2297. Retrieval augmentation does not improve this baseline, with Recall@10 of 31.99% and MRR of 0.1937. Performance is strongly affected by concept frequency: Recall@10 is 7.74% for concepts appearing in only one or two training notes versus 43.90% for concepts appearing in more than ten. In addition, 9.66% of test note-concept pairs contain concepts unseen during training. These findings show that local lexical context and terminology coverage are major determinants of recommendation quality in low-resource settings and provide a reproducible baseline for future ontology-grounded and biomedical-encoder retrieval systems.
Abstract (translated)
URL
https://arxiv.org/abs/2609.17855
PDF
-
Beyond Performance Metrics: Uncertainty Mapping of Label Ambiguity in Fazekas Score Prediction
2026-09-15 19:03:55Susanne Schmid, Johanna Ospel, Richard Frayne, Roberto SouzaAbstract
Reference labels used to train medical image classification models are not always as certain as they may appear, and this uncertainty has implications on performance metrics. In this study, we propose a framework to analyze model performance for periventricular Fazekas score prediction that goes beyond conventional metrics. The Fazekas score is an ordinal visual rating scale used to assess the severity of white matter hyperintensities and is known to be affected by inter-rater variability. While the best Fazekas score prediction model achieved a Matthews correlation coefficient (MCC) of 0.70, performance varied across data splits and loss functions, making interpretation of model capabilities difficult. Rather than interpreting epistemic uncertainty of a model's prediction as an isolated scalar value, our approach of uncertainty mapping relates uncertainty to its position within the learned feature representation. This highlights regions of class-boundary transitions where cases appear more ambiguous and misclassifications are more likely. It also identifies potential label disagreement, including low-uncertainty misclassified cases that expert review found to be inconsistent with the original reference Fazekas score. Therefore, uncertainty mapping allows model behaviour to be examined in relation to class separation and potential model-label disagreement. Loss function choice also influenced the uncertainty profile, with some models showing clearer class separation and more localized uncertainty in ambiguous regions than others. These findings suggest that uncertainty mapping for Fazekas score predictions can support model interpretation and targeted dataset review when reference labels are affected by ambiguity/ inter-rater variability.
Abstract (translated)
URL
https://arxiv.org/abs/2609.17753
PDF
-
Enhancing Accessibility of Medical Texts through Large Language Model-Driven Plain Language Adaptation
2026-09-15 16:33:38Ting-Wei Chang, Hen-Hsen Huang, Hsin-Hsi ChenAbstract
This paper addresses the challenge of making complex healthcare information more accessible through automated Plain Language Adaptation (PLA). PLA aims to simplify technical medical language, bridging a critical gap between the complexity of healthcare texts and patients' reading comprehension. Recent advances in Large Language Models (LLMs), such as GPT and BART, have opened new possibilities for PLA, especially in zero-shot and few-shot learning contexts where task-specific data is limited. In this work, we leverage the capabilities of LLMs such as GPT-4o-mini, Gemini-1.5-pro, and LLaMA for text simplification. Additionally, we incorporate Mixture-of-Agents (MoA) techniques to enhance adaptability and robustness in PLA tasks. Key contributions include a comparative analysis of prompting strategies, finetuning with QLoRA on different LLMs, and the integration of MoA technique. Our findings demonstrate the effectiveness of LLM-driven PLA, showcasing its potential in making healthcare information more comprehensible while preserving essential content.
Abstract (translated)
URL
https://arxiv.org/abs/2609.17398
PDF
-
MUMINS: Metadata-conditioned Uncertainty-aware Medical Image Next-state Synthesis
2026-09-15 13:32:37Anna Oliveras, Roger Mar\'i, Rafael Redondo, Oriol Guardi\`a, Cynthia Ifeyinwa Ugwu, Ana Tost, Bhalaji Nagarajan, Carolina Migliorelli, Vicent Ribas, Petia RadevaAbstract
Forecasting anatomical changes such as tumor growth and neurodegeneration is a challenging generative vision task. Morphological evolution is subtle relative to static anatomy, highly patient-specific, and inherently stochastic. Existing methods struggle with several issues: deterministic networks ignore biological stochasticity, while standard diffusion models require computationally prohibitive multi-pass sampling to quantify uncertainty. We propose MUMINS (Metadata-conditioned Uncertainty-aware Medical Image Next-state Synthesis), an efficient diffusion framework that jointly diffuses a baseline scan and its follow-up residual, summed to synthesize the follow-up scan, while concurrently predicting a spatial uncertainty map, in a single reverse diffusion process. Conditioned on the time interval and relevant metadata, it preserves fine-grained anatomy by dynamically re-injecting the baseline as a soft anchor at every denoising step, and a negative-log-likelihood head learns the uncertainty map to explicitly flag error-prone regions. Designed without organ-specific heuristics, the same architecture is reused across anatomies via separate, dataset-specific retraining. Extensive evaluations demonstrate that dataset-specific retraining of MUMINS matches or outperforms dedicated, domain-specific state-of-the-art methods on lung CT (PNG) and brain MRI (OASIS-3). Project page: this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2609.17169
PDF
-
Overcoming technical adoption barriers for mobile service robots in rehabilitation
2026-09-15 11:08:19Christian Sternitzke, Sebastian Blumenthal, Lukas Kleedoerfer, Verena Deserno, Anke MayfarthAbstract
Many publications on robotic systems in healthcare describe early-stage work on low technology readiness levels. This paper describes how a mobile service robot approved as a medical device reaches higher technology readiness levels by adding peripheral functions and smaller improvements, which are pivotal for user acceptance in clinical environments and which often cannot be elicited by questioning users ex-ante as certain aspects only come in mind from testing the systems in clinical settings or operational environments. Especially developers of service robots in healthcare are advised to plan with such downstream developments, which can take significant implementation time, to obtain user acceptance and achieve widespread adoption of their robotic systems.
Abstract (translated)
URL
https://arxiv.org/abs/2609.16996
PDF
-
MedPCFM-TED: One-Step Point Cloud Flow Matching for Implant Generation via Teacher-Guided Endpoint Distillation
2026-09-15 10:10:17Kamil Kwarciak, Marek WodzinskiAbstract
Cranial implant generation is an important task in medical imaging. Recent point cloud based generative methods, particularly flow matching, offer strong reconstruction quality and efficient sampling, but still require multiple neural function evaluations during inference. This limits rapid generation of multiple plausible implant candidates. We propose Teacher-guided Endpoint Distillation (TED), a simple one-step distillation framework for conditional cranial implant generation on point clouds. TED trains a one-step student using teacher-guided endpoint supervision and geometric matching losses, while avoiding explicit path straightening. We evaluate TED on the SkullFix and SkullBreak benchmarks. TED achieves the best overall performance on the SkullBreak dataset, remains competitive on SkullFix, and provides the strongest Chamfer distance performance among the compared one-step methods. In addition, TED generates implants in approximately 0.04s per sample. These results show that one-step distillation can substantially accelerate conditional point cloud implant generation without sacrificing reconstruction quality.
Abstract (translated)
URL
https://arxiv.org/abs/2609.16934
PDF
-
PiPS: Post-Hoc Prototypical Explanations for Interpretable Semantic Segmentation
2026-09-15 09:40:37Mi{\l}osz Adamczyk, Tymoteusz Zapala, Piotr Borycki, Przemys{\l}aw SpurekAbstract
With the increasing deployment of deep neural networks in critical systems, such as medical diagnostics and autonomous vehicles, ensuring their interpretability is crucial to building trust in decision-making systems. In the field of explainable artificial intelligence, prototype-based reasoning has gained particular popularity, as it mimics human cognitive processes by explaining model decisions based on visual similarity under the looks like this paradigm. While this paradigm has been thoroughly investigated in the context of global image classification, the interpretability of dense predictions, particularly semantic segmentation, remains largely unexplored despite its immense importance in tasks requiring precise object localization. Existing prototype-based interpretable segmentation models rely on ante-hoc architectures, which entails significant limitations because they require costly training from scratch and modifications to the network structure, ultimately leading to a noticeable drop in predictive performance compared to standard black-box models. To address this issue, we propose PiPS (Post-hoc interpretable Prototypical Segmentation), the first fully post-hoc solution for generating prototypical explanations for semantic segmentation models. Our method enables the extraction of intuitive, spatially localized explanations from any pre-trained network without modification or fine-tuning, thereby preserving 100% of the model's original predictive performance. This approach opens a new avenue for the safe and cost-effective deployment of transparent systems in advanced computer vision tasks. Codebase available at this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2609.16909
PDF
-
Smarter by the Moment: Environment-Driven Dynamic Policies for Continual LLM Improvement
2026-09-15 08:06:17Ting-Wei Chang, Po-Chun Chen, Hen-Hsen Huang, Hsin-Hsi ChenAbstract
Large Language Models (LLMs) have achieved remarkable progress across diverse domains, but continual adaptation to evolving tasks and environments remains a key challenge. Existing memory-augmented approaches retrieve individual past examples as direct references, but do not explicitly synthesize actionable strategies from them, causing the same types of errors to recur. We propose Dynamic Retrieval-based Policy Generation (DRPG), a framework that integrates memory-based retrieval with a dynamic policy generator, leveraging historical data and environment feedback to produce task-specific policies for continual LLM improvement. We evaluate DRPG across six benchmarks spanning text-to-SQL, question answering, medical diagnosis, and Python programming, using seven LLMs from both proprietary and open-weight families. DRPG outperforms strong baselines across most datasets and models. Further analysis demonstrates that DRPG's policy generation is robust to retrieval strategy, operates effectively without prior policy continuity, and can leverage smaller or cross-family models as cost-efficient policy generators. We also find that the benefit of policy-level guidance depends on task characteristics, offering practical insights into when and under what conditions this mechanism is most effective.
Abstract (translated)
URL
https://arxiv.org/abs/2609.16800
PDF
-
IMVS: Interactive Medical Volume Segmentation with Test-Time Adaptation - A New Method for Annotating Radiology Datasets
2026-09-15 07:43:38Abhilaksh Singh Reen, Kushal Borkar, Ritvik MahapatraAbstract
Annotating large radiology datasets is bottlenecked by the manual effort of delineating structures slice-by-slice in 3D volumes. Interactive methods reduce this effort but stay interaction-inefficient: slice-wise methods (including many foundation models) ignore inter-slice continuity, while 3D and video-based methods propagate a prompt with a \emph{fixed} propagator that never adapts to the target volume, so it drifts on low-contrast or pathological structures and must be re-prompted. We present IMVS, a human-in-the-loop annotation framework that composes three components into a closed loop rather than a new segmentation primitive: a lightweight 2D Slice Mask Adapter (SMA) fine-tuned online from user scribbles, a frozen Volume Mask Tracker (VMT) that propagates corrected masks across adjacent slices, and a soft teacher--student alignment that limits forgetting. The SMA is backbone-agnostic (UNet++, DeepLabV3, TransUNet). Across 8 public CT/MRI datasets, IMVS matches strong interactive baselines in quality while sharply cutting annotation effort: $14.4\times$ faster than a proficient copy-based manual workflow ($22.3\times$ over naive manual), $4.6\times$ over slice-wise and $1.9\times$ over 3D interactive methods. MedSAM2 and ScribblePrompt stay competitive or stronger on well-delineated organs; IMVS's advantage is largest on challenging targets and on interaction efficiency. Source code and Demo Video: this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2609.16775
PDF
-
De-GAN - Dynamic Parameter Tuned GAN for 3D Medical Image Segmentation: A Step Towards Generalisation
Abstract
Brain tumor segmentation remains difficult because enhancing tumor (ET) has low contrast and overlaps surrounding tissue, while scanner and site variation causes domain shift. We propose DE-GAN, a contrast-enhancing conditional GAN that combines input-adaptive dynamic convolutions, style-aware feature mixing, and coordinate encoding to synthesize slice-adaptive FLAIR images. A label-guided, class-conditional target separates tumor-core (TC) and ET intensities while preserving anatomy. The generated FLAIR is concatenated with the original MR modalities and used to train a 3D U-Net. Across BraTS 2015, 2018, and 2019, DE-GAN improves segmentation over the baseline and static EnhGAN replacement on most reported TC/ET metrics, with the largest gains from retaining both original and enhanced FLAIR. Code and pretrained models are available at this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2609.16755
PDF
-
TAME: Token Attribution and Masking for Emergent misalignment
2026-09-15 07:30:01Md Rayhanul Masud, Md Rizwan ParvezAbstract
Fine-tuning an aligned language model on narrow, flawed data can induce harmful behavior far outside the training domain, known as emergent misalignment (EM). Prior work has localized EM in model weights, activations, and training documents, but it remains unclear which training tokens carry the relevant fine-tuning signal. We introduce TAME (Token Attribution and Masking for Emergent Misalignment), a three-stage framework: token attribution scores how strongly the fine-tuning update raises each response token's likelihood, using forward passes through a released LoRA adapter; signal characterization finds patterns among high-attribution tokens; and causal validation tests them by attribution-guided loss masking. On released EM organisms and a 6,849-example medical-advice split, attribution is concentrated (the top 5% of tokens hold 32% of the mass) and, in Llama, depleted for medical vocabulary but enriched for a register of unwarranted certainty, even after controlling for token rarity. Masking high-attribution tokens during fresh fine-tuning cuts EM by 23x in Llama and 36x in Qwen, with the perplexity cost concentrated on the targeted register rather than on medical content; an equal random mask leaves EM unchanged. In Llama, the attribution pattern suggests that EM-relevant signal lies more in how confidently flawed content is expressed than in its domain vocabulary; the causal masking effect itself holds across both model families.
Abstract (translated)
URL
https://arxiv.org/abs/2609.16754
PDF
-
Japanese Stroke LLM Evaluation: A Conversational Benchmark for Safe Stroke Care in Japanese Using Large Language Models
2026-09-15 07:13:49Keisuke Masuda, Kazutaka Yatsushiro, Hirohumi Iwamoto, Hirofumi Hirano, Ryosuke HanayaAbstract
Background: Large language models (LLMs) have achieved physician-comparable performance on multiple-choice medical knowledge examinations, but their capabilities in clinical history taking, urgency assessment, and safety remain insufficiently evaluated. We proposed Japanese Stroke LLM Evaluation, a multi-turn conversational benchmark for stroke care in Japanese, and evaluated LLM performance and safety under practice-oriented conditions. Methods: We created 10 stroke and related-condition cases and evaluated LLMs in multi-turn Japanese conversations. The LLM acted as physician, while a board-certified neurosurgeon acted as simulated patient and evaluator. Each case comprised history-taking and action phases scored using pre-specified criteria. Errors that could directly threaten life were defined as critical mistakes. The safety threshold was at least 80% overall with zero critical mistakes. Eighteen models were evaluated in October 2025 and June 2026. Results: Claude Fable 5 achieved the highest score (87.4%) with zero critical mistakes, followed by Claude Opus 4.7 (80.3%) and GLM-5.2 (75.6%). Two leaders met the safety threshold. Eleven models made 17 critical mistakes, including failure to confirm laboratory results or blood glucose before t-PA, surgery before airway stabilization, omission of cervical vascular evaluation, and t-PA outside its indication. History-taking question count correlated with history-taking score (r = 0.648, p = 0.007). Conclusions: Japanese Stroke LLM Evaluation provides a benchmark for LLM performance under practice-oriented conditions, including a cap on history-taking questions. Cases and evaluations were created by neurosurgical specialists rather than using an LLM-as-judge approach. Performance improved across cloud-based and on-premise models in 2026, with some exceeding the safety threshold. Further evaluation using real-world cases is required.
Abstract (translated)
URL
https://arxiv.org/abs/2609.16739
PDF
-
Optimal Transport Metric Learning for Feature Alignment in Partially Supervised Segmentation
2026-09-15 06:37:29Dakini Mallam Garba, Salim Abdou DaouraAbstract
Multi-organ segmentation is often challenged by partially annotated datasets and domain shifts across different imaging sources. To address these limitations, we propose a two-stage learning framework that efficiently leverages partial supervision. In the first stage, the model learns from available annotations to produce accurate segmentations of annotated organs, establishing robust feature representations. In the second stage, we introduce learnable organ prototypes and a Sinkhorn-triplet loss to enforce organ-wise feature consistency across datasets. This encourages latent embeddings of the same organ to remain close, while increasing separation between different organs, even when annotations are missing. Our approach achieves performance comparable to state-of-the-art methods on the BTCV dataset, while remaining computationally efficient. By explicitly aligning feature distributions rather than relying solely on pseudo-labels, the framework effectively mitigates domain shift, making it particularly suitable for medical image segmentation tasks with limited annotation resources.
Abstract (translated)
URL
https://arxiv.org/abs/2609.19176
PDF
-
Rewarding Reasoning, Not Answers: Fixing and Bounding Test-Time Reinforcement Learning on Medical QA
2026-09-15 05:31:55Kailong Fan, Anqi Pu, Yichen Wu, Wanhua Li, Yicong Li, Hanspeter Pfister, Huafeng Liu, Xiang Li, Quanzheng Li, Ning GuoAbstract
Test-time reinforcement learning adapts a model on its own unlabeled test set using majority-vote pseudo-labels and has shown strong results in mathematics. We show that this recipe collapses on medical multiple-choice QA: accuracy stagnates while output diversity rapidly declines. Through a controlled experiment that keeps the questions, model, and optimizer fixed while changing only the answer space, we trace this failure to answer-space structure rather than domain difficulty. In small answer spaces, incorrect rollouts often collide on the same wrong pseudo-label and reinforce it; in large answer spaces, they disperse and receive little reward. This diagnosis motivates PROSE, Process Reward Guided Self-Training, which rewards reasoning quality instead of answer agreement. PROSE scores each reasoning step with a medical process reward model, assigns the trajectory reward as the minimum score across steps, and enforces answer-format constraints. Without labels, PROSE substantially improves a general Llama model, surpassing purpose-built medical models and matching much larger systems. Because the process signal is internalized into the policy, the adapted model requires no reward model at inference and transfers its gains to unseen datasets. We further show that the minimum aggregation is essential: mean aggregation can be exploited, saturating the proxy reward while degrading accuracy.
Abstract (translated)
URL
https://arxiv.org/abs/2609.16660
PDF
