Yan Hu

Heng Li, Haofeng Liu, Huazhu Fu et al.

Fundus photography is a routine examination in clinics to diagnose and monitor ocular diseases. However, for cataract patients, the fundus image always suffers quality degradation caused by the clouding lens. The degradation prevents reliable diagnosis by ophthalmologists or computer-aided systems. To improve the certainty in clinical diagnosis, restoration algorithms have been proposed to enhance the quality of fundus images. Unfortunately, challenges remain in the deployment of these algorithms, such as collecting sufficient training data and preserving retinal structures. In this paper, to circumvent the strict deployment requirement, a structure-consistent restoration network (SCR-Net) for cataract fundus images is developed from synthesized data that shares an identical structure. A cataract simulation model is firstly designed to collect synthesized cataract sets (SCS) formed by cataract fundus images sharing identical structures. Then high-frequency components (HFCs) are extracted from the SCS to constrain structure consistency such that the structure preservation in SCR-Net is enforced. The experiments demonstrate the effectiveness of SCR-Net in the comparison with state-of-the-art methods and the follow-up clinical applications. The code is available at https://github.com/liamheng/ArcNet-Medical-Image-Enhancement.

Heng Li, Haojin Li, Wei Zhao et al.

The annotation scarcity of medical image segmentation poses challenges in collecting sufficient training data for deep learning models. Specifically, models trained on limited data may not generalize well to other unseen data domains, resulting in a domain shift issue. Consequently, domain generalization (DG) is developed to boost the performance of segmentation models on unseen domains. However, the DG setup requires multiple source domains, which impedes the efficient deployment of segmentation algorithms in clinical scenarios. To address this challenge and improve the segmentation model's generalizability, we propose a novel approach called the Frequency-mixed Single-source Domain Generalization method (FreeSDG). By analyzing the frequency's effect on domain discrepancy, FreeSDG leverages a mixed frequency spectrum to augment the single-source domain. Additionally, self-supervision is constructed in the domain augmentation to learn robust context-aware representations for the segmentation task. Experimental results on five datasets of three modalities demonstrate the effectiveness of the proposed algorithm. FreeSDG outperforms state-of-the-art methods and significantly improves the segmentation model's generalizability. Therefore, FreeSDG provides a promising solution for enhancing the generalization of medical image segmentation models, especially when annotated data is scarce. The code is available at https://github.com/liamheng/Non-IID_Medical_Image_Segmentation.

Haofeng Liu, Heng Li, Huazhu Fu et al.

As an economical and efficient fundus imaging modality, retinal fundus images have been widely adopted in clinical fundus examination. Unfortunately, fundus images often suffer from quality degradation caused by imaging interferences, leading to misdiagnosis. Despite impressive enhancement performances that state-of-the-art methods have achieved, challenges remain in clinical scenarios. For boosting the clinical deployment of fundus image enhancement, this paper proposes the pyramid constraint to develop a degradation-invariant enhancement network (PCE-Net), which mitigates the demand for clinical data and stably enhances unknown data. Firstly, high-quality images are randomly degraded to form sequences of low-quality ones sharing the same content (SeqLCs). Then individual low-quality images are decomposed to Laplacian pyramid features (LPF) as the multi-level input for the enhancement. Subsequently, a feature pyramid constraint (FPC) for the sequence is introduced to enforce the PCE-Net to learn a degradation-invariant model. Extensive experiments have been conducted under the evaluation metrics of enhancement and segmentation. The effectiveness of the PCE-Net was demonstrated in comparison with state-of-the-art methods and the ablation study. The source code of this study is publicly available at https://github.com/HeverLaw/PCENet-Image-Enhancement.

Zhongxi Qiu, Yan Hu, Heng Li et al.

Segmentation is vital for ophthalmology image analysis. But its various modal images hinder most of the existing segmentation algorithms applications, as they rely on training based on a large number of labels or hold weak generalization ability. Based on Segment Anything (SAM), we propose a simple but effective learnable prompt layer suitable for multiple target segmentation in ophthalmology multi-modal images, named Learnable Ophthalmology Segment Anything (SAM). The learnable prompt layer learns medical prior knowledge from each transformer layer. During training, we only train the prompt layer and task head based on a one-shot mechanism. We demonstrate the effectiveness of our thought based on four medical segmentation tasks based on nine publicly available datasets. Moreover, we only provide a new improvement thought for applying the existing fundamental CV models in the medical field. Our codes are available at \href{https://github.com/Qsingle/LearnablePromptSAM}{website}.

Heng Li, Haofeng Liu, Yan Hu et al.

Cataracts are the leading cause of vision loss worldwide. Restoration algorithms are developed to improve the readability of cataract fundus images in order to increase the certainty in diagnosis and treatment for cataract patients. Unfortunately, the requirement of annotation limits the application of these algorithms in clinics. This paper proposes a network to annotation-freely restore cataractous fundus images (ArcNet) so as to boost the clinical practicability of restoration. Annotations are unnecessary in ArcNet, where the high-frequency component is extracted from fundus images to replace segmentation in the preservation of retinal structures. The restoration model is learned from the synthesized images and adapted to real cataract images. Extensive experiments are implemented to verify the performance and effectiveness of ArcNet. Favorable performance is achieved using ArcNet against state-of-the-art algorithms, and the diagnosis of ocular fundus diseases in cataract patients is promoted by ArcNet. The capability of properly restoring cataractous images in the absence of annotated data promises the proposed algorithm outstanding clinical practicability.

Cathy Shyr, Yan Hu, Paul A. Harris et al.

Rare diseases (RDs) are collectively common and affect 300 million people worldwide. Accurate phenotyping is critical for informing diagnosis and treatment, but RD phenotypes are often embedded in unstructured text and time-consuming to extract manually. While natural language processing (NLP) models can perform named entity recognition (NER) to automate extraction, a major bottleneck is the development of a large, annotated corpus for model training. Recently, prompt learning emerged as an NLP paradigm that can lead to more generalizable results without any (zero-shot) or few labeled samples (few-shot). Despite growing interest in ChatGPT, a revolutionary large language model capable of following complex human prompts and generating high-quality responses, none have studied its NER performance for RDs in the zero- and few-shot settings. To this end, we engineered novel prompts aimed at extracting RD phenotypes and, to the best of our knowledge, are the first the establish a benchmark for evaluating ChatGPT's performance in these settings. We compared its performance to the traditional fine-tuning approach and conducted an in-depth error analysis. Overall, fine-tuning BioClinicalBERT resulted in higher performance (F1 of 0.689) than ChatGPT (F1 of 0.472 and 0.591 in the zero- and few-shot settings, respectively). Despite this, ChatGPT achieved similar or higher accuracy for certain entities (i.e., rare diseases and signs) in the one-shot setting (F1 of 0.776 and 0.725). This suggests that with appropriate prompt engineering, ChatGPT has the potential to match or outperform fine-tuned language models for certain entity types with just one labeled sample. While the proliferation of large language models may provide opportunities for supporting RD diagnosis and treatment, researchers and clinicians should critically evaluate model outputs and be well-informed of their limitations.

Muhammad Asif Ali, Maha Alshmrani, Jianbin Qin et al.

Bilingual Lexical Induction (BLI) is a core challenge in NLP, it relies on the relative isomorphism of individual embedding spaces. Existing attempts aimed at controlling the relative isomorphism of different embedding spaces fail to incorporate the impact of semantically related words in the model training objective. To address this, we propose GARI that combines the distributional training objectives with multiple isomorphism losses guided by the graph attention network. GARI considers the impact of semantical variations of words in order to define the relative isomorphism of the embedding spaces. Experimental evaluation using the Arabic language data set shows that GARI outperforms the existing research by improving the average P@1 by a relative score of up to 40.95% and 76.80% for in-domain and domain mismatch settings respectively. We release the codes for GARI at https://github.com/asif6827/GARI.

Muhammad Asif Ali, Yan Hu, Jianbin Qin et al.

Automated construction of bilingual dictionaries using monolingual embedding spaces is a core challenge in machine translation. The end performance of these dictionaries relies upon the geometric similarity of individual spaces, i.e., their degree of isomorphism. Existing attempts aimed at controlling the relative isomorphism of different spaces fail to incorporate the impact of semantically related words in the training objective. To address this, we propose GRI that combines the distributional training objectives with attentive graph convolutions to unanimously consider the impact of semantically similar words required to define/compute the relative isomorphism of multiple spaces. Experimental evaluation shows that GRI outperforms the existing research by improving the average P@1 by a relative score of up to 63.6%. We release the codes for GRI at https://github.com/asif6827/GRI.

Dubai Li, Yuxiang He, Yan Hu et al.

Observational studies can yield clinically actionable evidence at scale, but executing them on real-world databases is open-ended and requires coherent decisions across cohort construction, analysis, and reporting. Prior evaluations of LLM agents emphasize isolated steps or single answers, missing the integrity and internal structure of the resulting evidence bundle. To address this gap, we introduce RWE-bench, a benchmark grounded in MIMIC-IV and derived from peer-reviewed observational studies. Each task provides the corresponding study protocol as the reference standard, requiring agents to execute experiments in a real database and iteratively generate tree-structured evidence bundles. We evaluate six LLMs (three open-source, three closed-source) under three agent scaffolds using both question-level correctness and end-to-end task metrics. Across 162 tasks, task success is low: the best agent reaches 39.9%, and the best open-source model reaches 30.4%. Agent scaffolds also matter substantially, causing over 30% variation in performance metrics. Furthermore, we implement an automated cohort evaluation method to rapidly localize errors and identify agent failure modes. Overall, the results highlight persistent limitations in agents' ability to produce end-to-end evidence bundles, and efficient validation remains an important direction for future work. Code and data are available at https://github.com/somewordstoolate/RWE-bench.

Xiaoqing Zhang, Zunjie Xiao, Xiao Wu et al.

Spatial attention mechanism has been widely incorporated into deep neural networks (DNNs), significantly lifting the performance in computer vision tasks via long-range dependency modeling. However, it may perform poorly in medical image analysis. Unfortunately, existing efforts are often unaware that long-range dependency modeling has limitations in highlighting subtle lesion regions. To overcome this limitation, we propose a practical yet lightweight architectural unit, Pyramid Pixel Context Adaption (PPCA) module, which exploits multi-scale pixel context information to recalibrate pixel position in a pixel-independent manner dynamically. PPCA first applies a well-designed cross-channel pyramid pooling to aggregate multi-scale pixel context information, then eliminates the inconsistency among them by the well-designed pixel normalization, and finally estimates per pixel attention weight via a pixel context integration. By embedding PPCA into a DNN with negligible overhead, the PPCANet is developed for medical image classification. In addition, we introduce supervised contrastive learning to enhance feature representation by exploiting the potential of label information via supervised contrastive loss. The extensive experiments on six medical image datasets show that PPCANet outperforms state-of-the-art attention-based networks and recent deep neural networks. We also provide visual analysis and ablation study to explain the behavior of PPCANet in the decision-making process.

Yan Hu, Zhongxi Qiu, Dan Zeng et al.

Vascular segmentation extracts blood vessels from images and serves as the basis for diagnosing various diseases, like ophthalmic diseases. Ophthalmologists often require high-resolution segmentation results for analysis, which leads to super-computational load by most existing methods. If based on low-resolution input, they easily ignore tiny vessels or cause discontinuity of segmented vessels. To solve these problems, the paper proposes an algorithm named SuperVessel, which gives out high-resolution and accurate vessel segmentation using low-resolution images as input. We first take super-resolution as our auxiliary branch to provide potential high-resolution detail features, which can be deleted in the test phase. Secondly, we propose two modules to enhance the features of the interested segmentation region, including an upsampling with feature decomposition (UFD) module and a feature interaction module (FIM) with a constraining loss to focus on the interested features. Extensive experiments on three publicly available datasets demonstrate that our proposed SuperVessel can segment more tiny vessels with higher segmentation accuracy IoU over 6%, compared with other state-of-the-art algorithms. Besides, the stability of SuperVessel is also stronger than other algorithms. We will release the code after the paper is published.

Jiayi Zhang, Xiaoshan Chen, Zhongxi Qiu et al.

Hard exudates (HE) is the most specific biomarker for retina edema. Precise HE segmentation is vital for disease diagnosis and treatment, but automatic segmentation is challenged by its large variation of characteristics including size, shape and position, which makes it difficult to detect tiny lesions and lesion boundaries. Considering the complementary features between segmentation and super-resolution tasks, this paper proposes a novel hard exudates segmentation method named SS-MAF with an auxiliary super-resolution task, which brings in helpful detailed features for tiny lesion and boundaries detection. Specifically, we propose a fusion module named Multi-scale Attention Fusion (MAF) module for our dual-stream framework to effectively integrate features of the two tasks. MAF first adopts split spatial convolutional (SSC) layer for multi-scale features extraction and then utilize attention mechanism for features fusion of the two tasks. Considering pixel dependency, we introduce region mutual information (RMI) loss to optimize MAF module for tiny lesions and boundary detection. We evaluate our method on two public lesion datasets, IDRiD and E-Ophtha. Our method shows competitive performance with low-resolution inputs, both quantitatively and qualitatively. On E-Ophtha dataset, the method can achieve $\geq3\%$ higher dice and recall compared with the state-of-the-art methods.

Anran Li, Yuanyuan Chen, Wenjun Long et al.

Large language models (LLMs) have demonstrated strong performance on medical benchmarks, including question answering and diagnosis. To enable their use in clinical settings, LLMs are typically further adapted through continued pretraining or post-training using clinical data. However, most medical LLMs are trained on data from a single institution, which faces limitations in generalizability and safety in heterogeneous systems. Federated learning (FL) is a promising solution for enabling collaborative model development across healthcare institutions. Yet applying FL to LLMs in medicine remains fundamentally limited. First, conventional FL requires transmitting the full model during each communication round, which becomes impractical for multi-billion-parameter LLMs given the limited computational resources. Second, many FL algorithms implicitly assume data homogeneity, whereas real-world clinical data are highly heterogeneous across patients, diseases, and institutional practices. We introduce the model-agnostic and parameter-efficient federated learning framework for adapting LLMs to medical applications. Fed-MedLoRA transmits only low-rank adapter parameters, reducing communication and computation overhead, while Fed-MedLoRA+ further incorporates adaptive, data-aware aggregation to improve convergence under cross-site heterogeneity. We apply the framework to clinical information extraction (IE), which transforms patient narratives into structured medical entities and relations. Accuracy was assessed across five patient cohorts through comparisons with BERT models, and LLaMA-3 and DeepSeek-R1, GPT-4o models. Evaluation settings included (1) in-domain training and testing, (2) external validation on independent cohorts, and (3) a low-resource new-site adaptation scenario using real-world clinical notes from the Yale New Haven Health System.

Zehan Li, Yan Hu, Scott Lane et al.

Accurate identification and categorization of suicidal events can yield better suicide precautions, reducing operational burden, and improving care quality in high-acuity psychiatric settings. Pre-trained language models offer promise for identifying suicidality from unstructured clinical narratives. We evaluated the performance of four BERT-based models using two fine-tuning strategies (multiple single-label and single multi-label) for detecting coexisting suicidal events from 500 annotated psychiatric evaluation notes. The notes were labeled for suicidal ideation (SI), suicide attempts (SA), exposure to suicide (ES), and non-suicidal self-injury (NSSI). RoBERTa outperformed other models using multiple single-label classification strategy (acc=0.86, F1=0.78). MentalBERT (acc=0.83, F1=0.74) also exceeded BioClinicalBERT (acc=0.82, F1=0.72) which outperformed BERT (acc=0.80, F1=0.70). RoBERTa fine-tuned with single multi-label classification further improved the model performance (acc=0.88, F1=0.81). The findings highlight that the model optimization, pretraining with domain-relevant data, and the single multi-label classification strategy enhance the model performance of suicide phenotyping. Keywords: EHR-based Phenotyping; Natural Language Processing; Secondary Use of EHR Data; Suicide Classification; BERT-based Model; Psychiatry; Mental Health

Chen Zhang, Qiuchi Li, Dawei Song et al.

Language model (LM) distillation aims at distilling the knowledge in a large teacher LM to a small student one. As a critical issue facing LM distillation, a superior student often arises from a teacher of a relatively small scale instead of a larger one, especially in the presence of substantial capacity gap between the teacher and student. This issue, often referred to as the \textit{curse of capacity gap}, suggests that there is likely an optimal teacher yielding the best-performing student along the scaling course of the teacher. Consequently, distillation trials on teachers of a wide range of scales are called for to determine the optimal teacher, which becomes computationally intensive in the context of large LMs (LLMs). This paper addresses this critical bottleneck by providing the \textit{law of capacity gap} inducted from a preliminary study on distilling a broad range of small-scale (<3B) LMs, where the optimal teacher consistently scales linearly with the student scale across different model and data scales. By extending the law to LLM distillation on a larger scale (7B), we succeed in obtaining versatile LLMs that outperform a wide array of competitors.

Hongyang Jiang, Mengdi Gao, Yan Hu et al.

Deep neural networks (DNNs) have been widely applied in medical image classification and achieve remarkable classification performance. These achievements heavily depend on large-scale accurately annotated training data. However, label noise is inevitably introduced in the medical image annotation, as the labeling process heavily relies on the expertise and experience of annotators. Meanwhile, DNNs suffer from overfitting noisy labels, degrading the performance of models. Therefore, in this work, we innovatively devise noise-robust training approach to mitigate the adverse effects of noisy labels in medical image classification. Specifically, we incorporate contrastive learning and intra-group attention mixup strategies into the vanilla supervised learning. The contrastive learning for feature extractor helps to enhance visual representation of DNNs. The intra-group attention mixup module constructs groups and assigns self-attention weights for group-wise samples, and subsequently interpolates massive noisy-suppressed samples through weighted mixup operation. We conduct comparative experiments on both synthetic and real-world noisy medical datasets under various noise levels. Rigorous experiments validate that our noise-robust method with contrastive learning and attention mixup can effectively handle with label noise, and is superior to state-of-the-art methods. An ablation study also shows that both components contribute to boost model performance. The proposed method demonstrates its capability of curb label noise and has certain potential toward real-world clinic applications.

Xidong Wang, Nuo Chen, Junyin Chen et al.

Despite the vast repository of global medical knowledge predominantly being in English, local languages are crucial for delivering tailored healthcare services, particularly in areas with limited medical resources. To extend the reach of medical AI advancements to a broader population, we aim to develop medical LLMs across the six most widely spoken languages, encompassing a global population of 6.1 billion. This effort culminates in the creation of the ApolloCorpora multilingual medical dataset and the XMedBench benchmark. In the multilingual medical benchmark, the released Apollo models, at various relatively-small sizes (i.e., 0.5B, 1.8B, 2B, 6B, and 7B), achieve the best performance among models of equivalent size. Especially, Apollo-7B is the state-of-the-art multilingual medical LLMs up to 70B. Additionally, these lite models could be used to improve the multi-lingual medical capabilities of larger models without fine-tuning in a proxy-tuning fashion. We will open-source training corpora, code, model weights and evaluation benchmark.

Cheng Guo, Fei Hu, Yan Hu

Passive millimeter-wave (PMMW) is a significant potential technique for human security screening. Several popular object detection networks have been used for PMMW images. However, restricted by the low resolution and high noise of PMMW images, PMMW hidden object detection based on deep learning usually suffers from low accuracy and low classification confidence. To tackle the above problems, this paper proposes a Task-Aligned Detection Transformer network, named PMMW-DETR. In the first stage, a Denoising Coarse-to-Fine Transformer (DCFT) backbone is designed to extract long- and short-range features in the different scales. In the second stage, we propose the Query Selection module to introduce learned spatial features into the network as prior knowledge, which enhances the semantic perception capability of the network. In the third stage, aiming to improve the classification performance, we perform a Task-Aligned Dual-Head block to decouple the classification and regression tasks. Based on our self-developed PMMW security screening dataset, experimental results including comparison with State-Of-The-Art (SOTA) methods and ablation study demonstrate that the PMMW-DETR obtains higher accuracy and classification confidence than previous works, and exhibits robustness to the PMMW images of low quality.

Xuhan Huang, Qingning Shen, Yan Hu et al.

Large Language Models (LLMs) have demonstrated strong performance across various natural language processing tasks, yet their proficiency in mathematical reasoning remains a key challenge. Addressing the gap between natural and mathematical language requires advanced reasoning capabilities, approaching those of Artificial General Intelligence (AGI). However, the evaluation remains challenging, as perfectly representing reality is inherently elusive, and traditional methods like manual or direct comparison of mathematical statements (Ramamonjison et al., 2023) are insufficient for assessing true modeling ability. We propose a process-oriented framework to evaluate LLMs' ability to construct mathematical models, using solvers to compare outputs with ground truth. Introducing Mamo, a benchmark with 1,209 questions covering ordinary differential equations, linear programming, and mixed-integer linear programming, we enable automatic evaluation of modeling accuracy. The results show that existing LLMs struggle with complex mathematical modeling tasks, with larger models demonstrating superior performance, while open-source models remain competitive in simpler cases but still fall short of proprietary models in more challenging problems.

Yan Hu, Qingyu Chen, Jingcheng Du et al.

Objective: This study quantifies the capabilities of GPT-3.5 and GPT-4 for clinical named entity recognition (NER) tasks and proposes task-specific prompts to improve their performance. Materials and Methods: We evaluated these models on two clinical NER tasks: (1) to extract medical problems, treatments, and tests from clinical notes in the MTSamples corpus, following the 2010 i2b2 concept extraction shared task, and (2) identifying nervous system disorder-related adverse events from safety reports in the vaccine adverse event reporting system (VAERS). To improve the GPT models' performance, we developed a clinical task-specific prompt framework that includes (1) baseline prompts with task description and format specification, (2) annotation guideline-based prompts, (3) error analysis-based instructions, and (4) annotated samples for few-shot learning. We assessed each prompt's effectiveness and compared the models to BioClinicalBERT. Results: Using baseline prompts, GPT-3.5 and GPT-4 achieved relaxed F1 scores of 0.634, 0.804 for MTSamples, and 0.301, 0.593 for VAERS. Additional prompt components consistently improved model performance. When all four components were used, GPT-3.5 and GPT-4 achieved relaxed F1 socres of 0.794, 0.861 for MTSamples and 0.676, 0.736 for VAERS, demonstrating the effectiveness of our prompt framework. Although these results trail BioClinicalBERT (F1 of 0.901 for the MTSamples dataset and 0.802 for the VAERS), it is very promising considering few training samples are needed. Conclusion: While direct application of GPT models to clinical NER tasks falls short of optimal performance, our task-specific prompt framework, incorporating medical knowledge and training samples, significantly enhances GPT models' feasibility for potential clinical applications.

Yan Hu, Xu Zuo, Yujia Zhou et al.

Backgrounds: Information extraction (IE) is critical in clinical natural language processing (NLP). While large language models (LLMs) excel on generative tasks, their performance on extractive tasks remains debated. Methods: We investigated Named Entity Recognition (NER) and Relation Extraction (RE) using 1,588 clinical notes from four sources (UT Physicians, MTSamples, MIMIC-III, and i2b2). We developed an annotated corpus covering 4 clinical entities and 16 modifiers, and compared instruction-tuned LLaMA-2 and LLaMA-3 against BERT in terms of performance, generalizability, computational resources, and throughput to BERT. Results: LLaMA models outperformed BERT across datasets. With sufficient training data, LLaMA showed modest improvements (1% on NER, 1.5-3.7% on RE); improvements were larger with limited training data. On unseen i2b2 data, LLaMA-3-70B outperformed BERT by 7% (F1) on NER and 4% on RE. However, LLaMA models required more computing resources and ran up to 28 times slower. We implemented "Kiwi," a clinical IE package featuring both models, available at https://kiwi.clinicalnlp.org/. Conclusion: This study is among the first to develop and evaluate a comprehensive clinical IE system using open-source LLMs. Results indicate that LLaMA models outperform BERT for clinical NER and RE but with higher computational costs and lower throughputs. These findings highlight that choosing between LLMs and traditional deep learning methods for clinical IE applications should remain task-specific, taking into account both performance metrics and practical considerations such as available computing resources and the intended use case scenarios.

Yuhao Du, Shunian Chen, Wenbo Zan et al.

The application of Large Language Models (LLMs) in Computer-Aided Design (CAD) remains an underexplored area, despite their remarkable advancements in other domains. In this paper, we present BlenderLLM, a novel framework for training LLMs specifically for CAD tasks leveraging a self-improvement methodology. To support this, we developed a bespoke training dataset, BlendNet, and introduced a comprehensive evaluation suite, CADBench. Our results reveal that existing models demonstrate significant limitations in generating accurate CAD scripts. However, through minimal instruction-based fine-tuning and iterative self-improvement, BlenderLLM significantly surpasses these models in both functionality and accuracy of CAD script generation. This research establishes a strong foundation for the application of LLMs in CAD while demonstrating the transformative potential of self-improving models in advancing CAD automation. We encourage further exploration and adoption of these methodologies to drive innovation in the field. The dataset, model, benchmark, and source code are publicly available at https://github.com/FreedomIntelligence/BlenderLLM

Luoying Hao, Yan Hu, Wenjun Lin et al.

Recognition and localization of surgical detailed actions is an essential component of developing a context-aware decision support system. However, most existing detection algorithms fail to provide high-accuracy action classes even having their locations, as they do not consider the surgery procedure's regularity in the whole video. This limitation hinders their application. Moreover, implementing the predictions in clinical applications seriously needs to convey model confidence to earn entrustment, which is unexplored in surgical action prediction. In this paper, to accurately detect fine-grained actions that happen at every moment, we propose an anchor-context action detection network (ACTNet), including an anchor-context detection (ACD) module and a class conditional diffusion (CCD) module, to answer the following questions: 1) where the actions happen; 2) what actions are; 3) how confidence predictions are. Specifically, the proposed ACD module spatially and temporally highlights the regions interacting with the extracted anchor in surgery video, which outputs action location and its class distribution based on anchor-context interactions. Considering the full distribution of action classes in videos, the CCD module adopts a denoising diffusion-based generative model conditioned on our ACD estimator to further reconstruct accurately the action predictions. Moreover, we utilize the stochastic nature of the diffusion model outputs to access model confidence for each prediction. Our method reports the state-of-the-art performance, with improvements of 4.0% mAP against baseline on the surgical video dataset.

Cathy Shyr, Yan Hu, Rory J. Tinker et al.

Phenotyping is fundamental to rare disease diagnosis, but manual curation of structured phenotypes from clinical notes is labor-intensive and difficult to scale. Existing artificial intelligence approaches typically optimize individual components of phenotyping but do not operationalize the full clinical workflow of extracting features from clinical text, standardizing them to Human Phenotype Ontology (HPO) terms, and prioritizing diagnostically informative HPO terms. We developed RARE-PHENIX, an end-to-end AI framework for rare disease phenotyping that integrates large language model-based phenotype extraction, ontology-grounded standardization to HPO terms, and supervised ranking of diagnostically informative phenotypes. We trained RARE-PHENIX using data from 2,671 patients across 11 Undiagnosed Diseases Network clinical sites, and externally validated it on 16,357 real-world clinical notes from Vanderbilt University Medical Center. Using clinician-curated HPO terms as the gold standard, RARE-PHENIX consistently outperformed a state-of-the-art deep learning baseline (PhenoBERT) across ontology-based similarity and precision-recall-F1 metrics in end-to-end evaluation (i.e., ontology-based similarity of 0.70 vs. 0.58). Ablation analyses demonstrated performance improvements with the addition of each module in RARE-PHENIX (extraction, standardization, and prioritization), supporting the value of modeling the full clinical phenotyping workflow. By modeling phenotyping as a clinically aligned workflow rather than a single extraction task, RARE-PHENIX provides structured, ranked phenotypes that are more concordant with clinician curation and has the potential to support human-in-the-loop rare disease diagnosis in real-world settings.

Luoying Hao, Yan Hu, Yang Yue et al.

A comprehensive and explicit understanding of surgical scenes plays a vital role in developing context-aware computer-assisted systems in the operating theatre. However, few works provide systematical analysis to enable hierarchical surgical scene understanding. In this work, we propose to represent the tasks set [phase recognition --> step recognition --> action and instrument detection] as multi-level semantic scene understanding (MSSU). For this target, we propose a novel hierarchical context transformer (HCT) network and thoroughly explore the relations across the different level tasks. Specifically, a hierarchical relation aggregation module (HRAM) is designed to concurrently relate entries inside multi-level interaction information and then augment task-specific features. To further boost the representation learning of the different tasks, inter-task contrastive learning (ICL) is presented to guide the model to learn task-wise features via absorbing complementary information from other tasks. Furthermore, considering the computational costs of the transformer, we propose HCT+ to integrate the spatial and temporal adapter to access competitive performance on substantially fewer tunable parameters. Extensive experiments on our cataract dataset and a publicly available endoscopic PSI-AVA dataset demonstrate the outstanding performance of our method, consistently exceeding the state-of-the-art methods by a large margin. The code is available at https://github.com/Aurora-hao/HCT.

Heng Li, Haojin Li, Jianyu Chen et al.

Deep learning models often encounter challenges in making accurate inferences when there are domain shifts between the source and target data. This issue is particularly pronounced in clinical settings due to the scarcity of annotated data resulting from the professional and private nature of medical data. Despite the existence of decent solutions, many of them are hindered in clinical settings due to limitations in data collection and computational complexity. To tackle domain shifts in data-scarce medical scenarios, we propose a Random frequency filtering enabled Single-source Domain Generalization algorithm (RaffeSDG), which promises robust out-of-domain inference with segmentation models trained on a single-source domain. A filter-based data augmentation strategy is first proposed to promote domain variability within a single-source domain by introducing variations in frequency space and blending homologous samples. Then Gaussian filter-based structural saliency is also leveraged to learn robust representations across augmented samples, further facilitating the training of generalizable segmentation models. To validate the effectiveness of RaffeSDG, we conducted extensive experiments involving out-of-domain inference on segmentation tasks for three human tissues imaged by four diverse modalities. Through thorough investigations and comparisons, compelling evidence was observed in these experiments, demonstrating the potential and generalizability of RaffeSDG. The code is available at https://github.com/liamheng/Non-IID_Medical_Image_Segmentation.

Qianqian Xie, Qingyu Chen, Aokun Chen et al.

Recent advancements in large language models (LLMs) like ChatGPT and LLaMA show promise in medical applications, yet challenges remain in medical language comprehension. This study presents Me-LLaMA, a new medical LLM family based on open-source LLaMA models, optimized for medical text analysis and diagnosis by leveraging large-scale, domain-specific datasets. The Me-LLaMA family, including foundation models Me-LLaMA 13/70B and their chat-enhanced versions, was developed through continued pre-training and instruction tuning with 129B tokens and 214K samples from biomedical and clinical sources. Training the 70B models required over 100,000 A100 GPU hours. Me-LLaMA's performance was evaluated across six medical text analysis tasks using 12 benchmark datasets and complex clinical case diagnosis, with automatic and human evaluations. Results indicate Me-LLaMA outperforms LLaMA and other open-source medical LLMs in zero-shot and supervised settings. Task-specific tuning further boosts performance, surpassing ChatGPT on 7 of 8 datasets and GPT-4 on 5 of 8. For complex clinical cases, Me-LLaMA achieves performance comparable to ChatGPT and GPT-4. This work underscores the importance of domain-specific data in developing medical LLMs and addresses the high computational costs involved in training, highlighting a balance between pre-training and fine-tuning strategies. Me-LLaMA models are now accessible under user agreements, providing a valuable resource for advancing medical AI.

Zhongxi Qiu, Zhang Zhang, Yan Hu et al.

This paper explores optimal data selection strategies for Reinforcement Learning with Verified Rewards (RLVR) training in the medical domain. While RLVR has shown exceptional potential for enhancing reasoning capabilities in large language models, most prior implementations have focused on mathematics and logical puzzles, with limited exploration of domain-specific applications like medicine. We investigate four distinct data sampling strategies from MedQA-USMLE: random sampling (baseline), and filtering using Phi-4, Gemma-3-27b-it, and Gemma-3-12b-it models. Using Gemma-3-12b-it as our base model and implementing Group Relative Policy Optimization (GRPO), we evaluate performance across multiple benchmarks including MMLU, GSM8K, MMLU-Pro, and CMMLU. Our findings demonstrate that models trained on filtered data generally outperform those trained on randomly selected samples. Notably, training on self-filtered samples (using Gemma-3-12b-it for filtering) achieved superior performance in medical domains but showed reduced robustness across different benchmarks, while filtering with larger models from the same series yielded better overall robustness. These results provide valuable insights into effective data organization strategies for RLVR in specialized domains and highlight the importance of thoughtful data selection in achieving optimal performance. You can access our repository (https://github.com/Qsingle/open-medical-r1) to get the codes.

Yan Hu, Ahmad Chaddad

This study introduces the SHAP-integrated convolutional diagnostic network (SICDN), an interpretable feature selection method designed for limited datasets, to address the challenge posed by data privacy regulations that restrict access to medical datasets. The SICDN model was tested on classification tasks using pneumonia and breast cancer datasets, demonstrating over 97% accuracy and surpassing four popular CNN models. We also integrated a historical weighted moving average technique to enhance feature selection. The SICDN shows potential in medical image prediction, with the code available on https://github.com/AIPMLab/SICDN.

Shuqi Yang, Mingrui Jing, Shuai Wang et al.

This study reviewed the use of Large Language Models (LLMs) in healthcare, focusing on their training corpora, customization techniques, and evaluation metrics. A systematic search of studies from 2021 to 2024 identified 61 articles. Four types of corpora were used: clinical resources, literature, open-source datasets, and web-crawled data. Common construction techniques included pre-training, prompt engineering, and retrieval-augmented generation, with 44 studies combining multiple methods. Evaluation metrics were categorized into process, usability, and outcome metrics, with outcome metrics divided into model-based and expert-assessed outcomes. The study identified critical gaps in corpus fairness, which contributed to biases from geographic, cultural, and socio-economic factors. The reliance on unverified or unstructured data highlighted the need for better integration of evidence-based clinical guidelines. Future research should focus on developing a tiered corpus architecture with vetted sources and dynamic weighting, while ensuring model transparency. Additionally, the lack of standardized evaluation frameworks for domain-specific models called for comprehensive validation of LLMs in real-world healthcare settings.

Anningzhe Gao, Shan Dai, Yan Hu

Irregular and asynchronous event sequences are prevalent in many domains, such as social media, finance, and healthcare. Traditional temporal point processes (TPPs), like Hawkes processes, often struggle to model mutual inhibition and nonlinearity effectively. While recent neural network models, including RNNs and Transformers, address some of these issues, they still face challenges with long-term dependencies and computational efficiency. In this paper, we introduce the Mamba Hawkes Process (MHP), which leverages the Mamba state space architecture to capture long-range dependencies and dynamic event interactions. Our results show that MHP outperforms existing models across various datasets. Additionally, we propose the Mamba Hawkes Process Extension (MHP-E), which combines Mamba and Transformer models to enhance predictive capabilities. We present the novel application of the Mamba architecture to Hawkes processes, a flexible and extensible model structure, and a theoretical analysis of the synergy between state space models and Hawkes processes. Experimental results demonstrate the superior performance of both MHP and MHP-E, advancing the field of temporal point process modeling.

Veronica Thai, Rui Li, Meng Ling et al.

Interpretation of giga-pixel whole-slide images (WSIs) is an important but difficult task for pathologists. Their diagnostic accuracy is estimated to average around 70%. Adding a second pathologist does not substantially improve decision consistency. The field lacks adequate behavioral data to explain diagnostic errors and inconsistencies. To fill in this gap, we present PathoGaze1.0, a comprehensive behavioral dataset capturing the dynamic visual search and decision-making processes of the full diagnostic workflow during cancer diagnosis. The dataset comprises 18.69 hours of eye-tracking, mouse interaction, stimulus tracking, viewport navigation, and diagnostic decision data (EMSVD) collected from 19 pathologists interpreting 397 WSIs. The data collection process emphasizes ecological validity through an application-grounded testbed, called PTAH. In total, we recorded 171,909 fixations, 263,320 saccades, and 1,867,362 mouse interaction events. In addition, such data could also be used to improve the training of both pathologists and AI systems that might support human experts. All experiments were preregistered at https://osf.io/hj9a7, and the complete dataset along with analysis code is available at https://go.osu.edu/pathogaze.

Zhengzhao Lai, Youbin Zheng, Zhenyang Cai et al.

Materials characterization is fundamental to acquiring materials information, revealing the processing-microstructure-property relationships that guide material design and optimization. While multimodal large language models (MLLMs) have recently shown promise in generative and predictive tasks within materials science, their capacity to understand real-world characterization imaging data remains underexplored. To bridge this gap, we present MatCha, the first benchmark for materials characterization image understanding, comprising 1,500 questions that demand expert-level domain expertise. MatCha encompasses four key stages of materials research comprising 21 distinct tasks, each designed to reflect authentic challenges faced by materials scientists. Our evaluation of state-of-the-art MLLMs on MatCha reveals a significant performance gap compared to human experts. These models exhibit degradation when addressing questions requiring higher-level expertise and sophisticated visual perception. Simple few-shot and chain-of-thought prompting struggle to alleviate these limitations. These findings highlight that existing MLLMs still exhibit limited adaptability to real-world materials characterization scenarios. We hope MatCha will facilitate future research in areas such as new material discovery and autonomous scientific agents. MatCha is available at https://github.com/FreedomIntelligence/MatCha.

Muhammad Asif Ali, Yan Hu, Jianbin Qin et al.

Antonyms vs synonyms distinction is a core challenge in lexico-semantic analysis and automated lexical resource construction. These pairs share a similar distributional context which makes it harder to distinguish them. Leading research in this regard attempts to capture the properties of the relation pairs, i.e., symmetry, transitivity, and trans-transitivity. However, the inability of existing research to appropriately model the relation-specific properties limits their end performance. In this paper, we propose InterlaCed Encoder NETworks (i.e., ICE-NET) for antonym vs synonym distinction, that aim to capture and model the relation-specific properties of the antonyms and synonyms pairs in order to perform the classification task in a performance-enhanced manner. Experimental evaluation using the benchmark datasets shows that ICE-NET outperforms the existing research by a relative score of upto 1.8% in F1-measure. We release the codes for ICE-NET at https://github.com/asif6827/ICENET.

Heng Li, Ziqin Lin, Zhongxi Qiu et al.

Medical imaging provides many valuable clues involving anatomical structure and pathological characteristics. However, image degradation is a common issue in clinical practice, which can adversely impact the observation and diagnosis by physicians and algorithms. Although extensive enhancement models have been developed, these models require a well pre-training before deployment, while failing to take advantage of the potential value of inference data after deployment. In this paper, we raise an algorithm for source-free unsupervised domain adaptive medical image enhancement (SAME), which adapts and optimizes enhancement models using test data in the inference phase. A structure-preserving enhancement network is first constructed to learn a robust source model from synthesized training data. Then a teacher-student model is initialized with the source model and conducts source-free unsupervised domain adaptation (SFUDA) by knowledge distillation with the test data. Additionally, a pseudo-label picker is developed to boost the knowledge distillation of enhancement tasks. Experiments were implemented on ten datasets from three medical image modalities to validate the advantage of the proposed algorithm, and setting analysis and ablation studies were also carried out to interpret the effectiveness of SAME. The remarkable enhancement performance and benefits for downstream tasks demonstrate the potential and generalizability of SAME. The code is available at https://github.com/liamheng/Annotation-free-Medical-Image-Enhancement.

Qingyu Chen, Yan Hu, Xueqing Peng et al.

The rapid growth of biomedical literature poses challenges for manual knowledge curation and synthesis. Biomedical Natural Language Processing (BioNLP) automates the process. While Large Language Models (LLMs) have shown promise in general domains, their effectiveness in BioNLP tasks remains unclear due to limited benchmarks and practical guidelines. We perform a systematic evaluation of four LLMs, GPT and LLaMA representatives on 12 BioNLP benchmarks across six applications. We compare their zero-shot, few-shot, and fine-tuning performance with traditional fine-tuning of BERT or BART models. We examine inconsistencies, missing information, hallucinations, and perform cost analysis. Here we show that traditional fine-tuning outperforms zero or few shot LLMs in most tasks. However, closed-source LLMs like GPT-4 excel in reasoning-related tasks such as medical question answering. Open source LLMs still require fine-tuning to close performance gaps. We find issues like missing information and hallucinations in LLM outputs. These results offer practical insights for applying LLMs in BioNLP.

Jifeng Xuan, He Jiang, Yan Hu et al.

Software companies spend over 45 percent of cost in dealing with software bugs. An inevitable step of fixing bugs is bug triage, which aims to correctly assign a developer to a new bug. To decrease the time cost in manual work, text classification techniques are applied to conduct automatic bug triage. In this paper, we address the problem of data reduction for bug triage, i.e., how to reduce the scale and improve the quality of bug data. We combine instance selection with feature selection to simultaneously reduce data scale on the bug dimension and the word dimension. To determine the order of applying instance selection and feature selection, we extract attributes from historical bug data sets and build a predictive model for a new bug data set. We empirically investigate the performance of data reduction on totally 600,000 bug reports of two large open source projects, namely Eclipse and Mozilla. The results show that our data reduction can effectively reduce the data scale and improve the accuracy of bug triage. Our work provides an approach to leveraging techniques on data processing to form reduced and high-quality bug data in software development and maintenance.

Xudong Lu, Yinghao Chen, Cheng Chen et al.

The emergence and growing popularity of multimodal large language models (MLLMs) have significant potential to enhance various aspects of daily life, from improving communication to facilitating learning and problem-solving. Mobile phones, as essential daily companions, represent the most effective and accessible deployment platform for MLLMs, enabling seamless integration into everyday tasks. However, deploying MLLMs on mobile phones presents challenges due to limitations in memory size and computational capability, making it difficult to achieve smooth and real-time processing without extensive optimization. In this paper, we present BlueLM-V-3B, an algorithm and system co-design approach specifically tailored for the efficient deployment of MLLMs on mobile platforms. To be specific, we redesign the dynamic resolution scheme adopted by mainstream MLLMs and implement system optimization for hardware-aware deployment to optimize model inference on mobile phones. BlueLM-V-3B boasts the following key highlights: (1) Small Size: BlueLM-V-3B features a language model with 2.7B parameters and a vision encoder with 400M parameters. (2) Fast Speed: BlueLM-V-3B achieves a generation speed of 24.4 token/s on the MediaTek Dimensity 9300 processor with 4-bit LLM weight quantization. (3) Strong Performance: BlueLM-V-3B has attained the highest average score of 66.1 on the OpenCompass benchmark among models with $\leq$ 4B parameters and surpassed a series of models with much larger parameter sizes (e.g., MiniCPM-V-2.6, InternVL2-8B).

Ahmad Chaddad, Yan Hu, Yihang Wu et al.

Objective. This paper presents an overview of generalizable and explainable artificial intelligence (XAI) in deep learning (DL) for medical imaging, aimed at addressing the urgent need for transparency and explainability in clinical applications. Methodology. We propose to use four CNNs in three medical datasets (brain tumor, skin cancer, and chest x-ray) for medical image classification tasks. In addition, we perform paired t-tests to show the significance of the differences observed between different methods. Furthermore, we propose to combine ResNet50 with five common XAI techniques to obtain explainable results for model prediction, aiming at improving model transparency. We also involve a quantitative metric (confidence increase) to evaluate the usefulness of XAI techniques. Key findings. The experimental results indicate that ResNet50 can achieve feasible accuracy and F1 score in all datasets (e.g., 86.31\% accuracy in skin cancer). Furthermore, the findings show that while certain XAI methods, such as XgradCAM, effectively highlight relevant abnormal regions in medical images, others, like EigenGradCAM, may perform less effectively in specific scenarios. In addition, XgradCAM indicates higher confidence increase (e.g., 0.12 in glioma tumor) compared to GradCAM++ (0.09) and LayerCAM (0.08). Implications. Based on the experimental results and recent advancements, we outline future research directions to enhance the robustness and generalizability of DL models in the field of biomedical imaging.

Haili Ye, Xiaoqing Zhang, Yan Hu et al.

The morphologies of vessel-like structures, such as blood vessels and nerve fibres, play significant roles in disease diagnosis, e.g., Parkinson's disease. Deep network-based refinement segmentation methods have recently achieved promising vessel-like structure segmentation results. There are still two challenges: (1) existing methods have limitations in rehabilitating subsection ruptures in segmented vessel-like structures; (2) they are often overconfident in predicted segmentation results. To tackle these two challenges, this paper attempts to leverage the potential of spatial interconnection relationships among subsection ruptures from the structure rehabilitation perspective. Based on this, we propose a novel Vessel-like Structure Rehabilitation Network (VSR-Net) to rehabilitate subsection ruptures and improve the model calibration based on coarse vessel-like structure segmentation results. VSR-Net first constructs subsection rupture clusters with Curvilinear Clustering Module (CCM). Then, the well-designed Curvilinear Merging Module (CMM) is applied to rehabilitate the subsection ruptures to obtain the refined vessel-like structures. Extensive experiments on five 2D/3D medical image datasets show that VSR-Net significantly outperforms state-of-the-art (SOTA) refinement segmentation methods with lower calibration error. Additionally, we provide quantitative analysis to explain the morphological difference between the rehabilitation results of VSR-Net and ground truth (GT), which is smaller than SOTA methods and GT, demonstrating that our method better rehabilitates vessel-like structures by restoring subsection ruptures.

Wenjun Lin, Yan Hu, Huazhu Fu et al.

Instrument-tissue interaction detection task, which helps understand surgical activities, is vital for constructing computer-assisted surgery systems but with many challenges. Firstly, most models represent instrument-tissue interaction in a coarse-grained way which only focuses on classification and lacks the ability to automatically detect instruments and tissues. Secondly, existing works do not fully consider relations between intra- and inter-frame of instruments and tissues. In the paper, we propose to represent instrument-tissue interaction as <instrument class, instrument bounding box, tissue class, tissue bounding box, action class> quintuple and present an Instrument-Tissue Interaction Detection Network (ITIDNet) to detect the quintuple for surgery videos understanding. Specifically, we propose a Snippet Consecutive Feature (SCF) Layer to enhance features by modeling relationships of proposals in the current frame using global context information in the video snippet. We also propose a Spatial Corresponding Attention (SCA) Layer to incorporate features of proposals between adjacent frames through spatial encoding. To reason relationships between instruments and tissues, a Temporal Graph (TG) Layer is proposed with intra-frame connections to exploit relationships between instruments and tissues in the same frame and inter-frame connections to model the temporal information for the same instance. For evaluation, we build a cataract surgery video (PhacoQ) dataset and a cholecystectomy surgery video (CholecQ) dataset. Experimental results demonstrate the promising performance of our model, which outperforms other state-of-the-art models on both datasets.

Xu Wang, Yan Hu, Wenyu Du et al.

Fine-tuning significantly improves the performance of Large Language Models (LLMs), yet its underlying mechanisms remain poorly understood. This paper aims to provide an in-depth interpretation of the fine-tuning process through circuit analysis, a popular tool in Mechanistic Interpretability (MI). Unlike previous studies (Prakash et al. 2024; Chhabra et al. 2024) that focus on tasks where pre-trained models already perform well, we develop a set of mathematical tasks where fine-tuning yields substantial performance gains, which are closer to the practical setting. In our experiments, we identify circuits at various checkpoints during fine-tuning and examine the interplay between circuit analysis, fine-tuning methods, and task complexities. First, we find that while circuits maintain high node similarity before and after fine-tuning, their edges undergo significant changes, in contrast to prior work that shows circuits only add some additional components after fine-tuning. Based on these observations, we develop a circuit-aware Low-Rank Adaptation (LoRA) method, which assigns ranks to layers based on edge changes in the circuits. Experimental results demonstrate that our circuit-based LoRA algorithm achieves an average performance improvement of 2.46% over standard LoRA with similar parameter sizes. Furthermore, we explore how combining circuits from subtasks can enhance fine-tuning in compositional tasks, providing new insights into the design of such tasks and deepening the understanding of circuit dynamics and fine-tuning mechanisms.

Qiushi Nie, Xiaoqing Zhang, Yan Hu et al.

Medical image registration is vital for disease diagnosis and treatment with its ability to merge diverse information of images, which may be captured under different times, angles, or modalities. Although several surveys have reviewed the development of medical image registration, these surveys have not systematically summarized methodologies of existing medical image registration methods. To this end, we provide a comprehensive review of these methods from traditional and deep learning-based directions, aiming to help audiences understand the development of medical image registration quickly. In particular, we review recent advances in retinal image registration at the end of each section, which has not attracted much attention. Additionally, we also discuss the current challenges of retinal image registration and provide insights and prospects for future research.

Wenjun Hou, Yi Cheng, Kaishuai Xu et al.

Previous research on radiology report generation has made significant progress in terms of increasing the clinical accuracy of generated reports. In this paper, we emphasize another crucial quality that it should possess, i.e., inter-report consistency, which refers to the capability of generating consistent reports for semantically equivalent radiographs. This quality is even of greater significance than the overall report accuracy in terms of ensuring the system's credibility, as a system prone to providing conflicting results would severely erode users' trust. Regrettably, existing approaches struggle to maintain inter-report consistency, exhibiting biases towards common patterns and susceptibility to lesion variants. To address this issue, we propose ICON, which improves the inter-report consistency of radiology report generation. Aiming to enhance the system's ability to capture similarities in semantically equivalent lesions, our approach first involves extracting lesions from input images and examining their characteristics. Then, we introduce a lesion-aware mixup technique to ensure that the representations of the semantically equivalent lesions align with the same attributes, achieved through a linear combination during the training phase. Extensive experiments on three publicly available chest X-ray datasets verify the effectiveness of our approach, both in terms of improving the consistency and accuracy of the generated reports.

Wenjun Hou, Yi Cheng, Kaishuai Xu et al.

Large language models (LLMs) have demonstrated remarkable capabilities in various domains, including radiology report generation. Previous approaches have attempted to utilize multimodal LLMs for this task, enhancing their performance through the integration of domain-specific knowledge retrieval. However, these approaches often overlook the knowledge already embedded within the LLMs, leading to redundant information integration. To address this limitation, we propose Radar, a framework for enhancing radiology report generation with supplementary knowledge injection. Radar improves report generation by systematically leveraging both the internal knowledge of an LLM and externally retrieved information. Specifically, it first extracts the model's acquired knowledge that aligns with expert image-based classification outputs. It then retrieves relevant supplementary knowledge to further enrich this information. Finally, by aggregating both sources, Radar generates more accurate and informative radiology reports. Extensive experiments on MIMIC-CXR, CheXpert-Plus, and IU X-ray demonstrate that our model outperforms state-of-the-art LLMs in both language quality and clinical accuracy.

Wenjun Hou, Yi Cheng, Kaishuai Xu et al.

Comprehensively understanding surgical scenes in Surgical Visual Question Answering (Surgical VQA) requires reasoning over multiple objects. Previous approaches address this task using cross-modal fusion strategies to enhance reasoning ability. However, these methods often struggle with limited scene understanding and question comprehension, and some rely on external resources (e.g., pre-extracted object features), which can introduce errors and generalize poorly across diverse surgical environments. To address these challenges, we propose SCAN, a simple yet effective memory-augmented framework that leverages Multimodal LLMs to improve surgical context comprehension via Self-Contained Inquiry. SCAN operates autonomously, generating two types of memory for context augmentation: Direct Memory (DM), which provides multiple candidates (or hints) to the final answer, and Indirect Memory (IM), which consists of self-contained question-hint pairs to capture broader scene context. DM directly assists in answering the question, while IM enhances understanding of the surgical scene beyond the immediate query. Reasoning over these object-aware memories enables the model to accurately interpret images and respond to questions. Extensive experiments on three publicly available Surgical VQA datasets demonstrate that SCAN achieves state-of-the-art performance, offering improved accuracy and robustness across various surgical scenarios.

Xu Wang, Zihao Li, Benyou Wang et al.

Large language models (LLMs) store vast amounts of information, making them powerful yet raising privacy and safety concerns when selective knowledge removal is required. Existing unlearning strategies, ranging from gradient-based fine-tuning and model editing to sparse autoencoder (SAE) steering, either lack interpretability or fail to provide a robust defense against adversarial prompts. We propose SAE-Guided Subspace Projection Unlearning (SSPU), a novel framework that leverages SAE features to drive targeted updates in the model's parameter space, enabling precise, interpretable, and robust unlearning. SSPU's three-stage pipeline performs data-driven layer and feature selection, subspace construction via QR decomposition, and constrained optimization that controls activations into an "irrelevant" subspace while preserving retained knowledge. Overall, we use SAE features to construct a subspace that supervises unlearning, refining the loss and adding a regularization term to guide interpretable parameter updates. In experiments on the WMDP-Cyber forget set and three utility benchmarks (MMLU, TruthfulQA, GSM8K), SSPU reduces harmful knowledge accuracy by 3.22% compared to the strongest baseline. It also improves adversarial robustness, lowering malicious accuracy under jailbreak prompts compared to baselines. Our findings expose the limitations of prior unlearning methods and demonstrate how interpretable subspace-guided optimization can achieve robust, controllable model behavior.

Baojiao Xiong, Boheng Chen, Chengzhi Wang et al. · baidu, tencent-ai

We present BlueLM-2.5-3B, a compact and unified dense Multimodal Large Language Model (MLLM) designed for efficient edge-device deployment, offering strong general-purpose and reasoning capabilities. To the best of our knowledge, this is the first 3B-scale MLLM to support both thinking and non-thinking modes, while also enabling explicit control over thinking token budget. BlueLM-2.5-3B is developed through diversified data curation, key data resampling, hybrid heterogeneous reinforcement learning, and a high-performance training infrastructure. Our model achieves superior multimodal capacity while preserving competitive pure-text performance with only 2.9 billion parameters. We conduct comprehensive evaluations across a broad range of multimodal and text-only benchmarks. In thinking mode, BlueLM-2.5-3B achieves comparable performance to Qwen3-4B on text-only benchmarks, and trails the larger Kimi-VL-A3B-16B by only about 5% on average across multimodal evaluations. In non-thinking mode, it outperforms Qwen2.5-VL-3B on the majority of multimodal benchmarks. Additionally, BlueLM-2.5-3B exhibits exceptional data efficiency. All of the aforementioned performance is achieved with substantially less total training data than Qwen2.5-VL-3B and Qwen3-4B. We hope our work contributes to the advancement of high-performance, on-device MLLMs and provides meaningful insights to the research community.

Yuzhe Yang, Yifei Zhang, Yan Hu et al.

This paper introduces the UCFE: User-Centric Financial Expertise benchmark, an innovative framework designed to evaluate the ability of large language models (LLMs) to handle complex real-world financial tasks. UCFE benchmark adopts a hybrid approach that combines human expert evaluations with dynamic, task-specific interactions to simulate the complexities of evolving financial scenarios. Firstly, we conducted a user study involving 804 participants, collecting their feedback on financial tasks. Secondly, based on this feedback, we created our dataset that encompasses a wide range of user intents and interactions. This dataset serves as the foundation for benchmarking 11 LLMs services using the LLM-as-Judge methodology. Our results show a significant alignment between benchmark scores and human preferences, with a Pearson correlation coefficient of 0.78, confirming the effectiveness of the UCFE dataset and our evaluation approach. UCFE benchmark not only reveals the potential of LLMs in the financial domain but also provides a robust framework for assessing their performance and user satisfaction.

Xiaoqing Zhang, Hanfeng Shi, Xiangyu Li et al.

Parkinson's disease (PD) is a prevalent neurodegenerative disorder globally. The eye's retina is an extension of the brain and has great potential in PD screening. Recent studies have suggested that texture features extracted from retinal layers can be adopted as biomarkers for PD diagnosis under optical coherence tomography (OCT) images. Frequency domain learning techniques can enhance the feature representations of deep neural networks (DNNs) by decomposing frequency components involving rich texture features. Additionally, previous works have not exploited texture features for automated PD screening in OCT. Motivated by the above analysis, we propose a novel Adaptive Wavelet Filter (AWF) that serves as the Practical Texture Feature Amplifier to fully leverage the merits of texture features to boost the PD screening performance of DNNs with the aid of frequency domain learning. Specifically, AWF first enhances texture feature representation diversities via channel mixer, then emphasizes informative texture feature representations with the well-designed adaptive wavelet filtering token mixer. By combining the AWFs with the DNN stem, AWFNet is constructed for automated PD screening. Additionally, we introduce a novel Balanced Confidence (BC) Loss by mining the potential of sample-wise predicted probabilities of all classes and class frequency prior, to further boost the PD screening performance and trustworthiness of AWFNet. The extensive experiments manifest the superiority of our AWFNet and BC over state-of-the-art methods in terms of PD screening performance and trustworthiness.