Hejie Cui

Yue Yu, Xuan Kan, Hejie Cui et al. · cmu

Functional magnetic resonance imaging (fMRI) has become one of the most common imaging modalities for brain function analysis. Recently, graph neural networks (GNN) have been adopted for fMRI analysis with superior performance. Unfortunately, traditional functional brain networks are mainly constructed based on similarities among region of interests (ROI), which are noisy and agnostic to the downstream prediction tasks and can lead to inferior results for GNN-based models. To better adapt GNNs for fMRI analysis, we propose TBDS, an end-to-end framework based on \underline{T}ask-aware \underline{B}rain connectivity \underline{D}AG (short for Directed Acyclic Graph) \underline{S}tructure generation for fMRI analysis. The key component of TBDS is the brain network generator which adopts a DAG learning approach to transform the raw time-series into task-aware brain connectivities. Besides, we design an additional contrastive regularization to inject task-specific knowledge during the brain network generation process. Comprehensive experiments on two fMRI datasets, namely Adolescent Brain Cognitive Development (ABCD) and Philadelphia Neuroimaging Cohort (PNC) datasets demonstrate the efficacy of TBDS. In addition, the generated brain networks also highlight the prediction-related brain regions and thus provide unique interpretations of the prediction results. Our implementation will be published to https://github.com/yueyu1030/TBDS upon acceptance.

Xuan Kan, Wei Dai, Hejie Cui et al.

Human brains are commonly modeled as networks of Regions of Interest (ROIs) and their connections for the understanding of brain functions and mental disorders. Recently, Transformer-based models have been studied over different types of data, including graphs, shown to bring performance gains widely. In this work, we study Transformer-based models for brain network analysis. Driven by the unique properties of data, we model brain networks as graphs with nodes of fixed size and order, which allows us to (1) use connection profiles as node features to provide natural and low-cost positional information and (2) learn pair-wise connection strengths among ROIs with efficient attention weights across individuals that are predictive towards downstream analysis tasks. Moreover, we propose an Orthonormal Clustering Readout operation based on self-supervised soft clustering and orthonormal projection. This design accounts for the underlying functional modules that determine similar behaviors among groups of ROIs, leading to distinguishable cluster-aware node embeddings and informative graph embeddings. Finally, we re-standardize the evaluation pipeline on the only one publicly available large-scale brain network dataset of ABIDE, to enable meaningful comparison of different models. Experiment results show clear improvements of our proposed Brain Network Transformer on both the public ABIDE and our restricted ABCD datasets. The implementation is available at https://github.com/Wayfear/BrainNetworkTransformer.

Yuzhang Xie, Keqi Han, Yunpeng Xiao et al.

Clinical decision-making (CDM) is central to real-world clinical workflows, where clinicians infer diagnoses, select treatments, or anticipate future health outcomes under incomplete evidence. LLMs are increasingly used to support these decisions due to strong language capabilities, broad biomedical knowledge, and efficiency, yet the reliability of LLMs on real-world clinical decision tasks remains insufficiently understood. To evaluate CDM models, especially LLM-based models, an ideal and practical medical decision benchmark should be constructed via an automated yet reliable pipeline to ensure both scale and quality. Moreover, the grounding of a CDM benchmark in real patient EHRs can better support evaluation on practical CDM tasks that require substantive biomedical knowledge and clinical inference. To fill the gaps, we introduce EHRBench, an automated and reliable EHR-grounded benchmark for evaluating LLM-based clinical decision-making at scale. To ensure scalability and reliability, EHRBench is constructed through an EHR-LLM-KB(knowledge-base) interaction pipeline. For efficiency, we use a specialized LLM to automatically convert encounter-level EHR trajectories into structured templates and deterministically instantiate the templates into QA items. In parallel, we apply systematic KB-based verification and enrichment to filter hallucinated or ambiguous relations and to improve reliability. Using this pipeline, we construct nearly 1M (960,067) QA items spanning three core inference-required clinical decision tasks: diagnosis, treatment, and prognosis. We benchmark more than 30 representative LLMs on EHRBench and provide detailed analyses of performance and robustness. The results show consistent capability trends across settings, further validating the reliability of EHRBench and highlighting actionable gaps toward clinically reliable LLM systems.

Xuan Kan, Hejie Cui, Joshua Lukemire et al.

Functional magnetic resonance imaging (fMRI) is one of the most common imaging modalities to investigate brain functions. Recent studies in neuroscience stress the great potential of functional brain networks constructed from fMRI data for clinical predictions. Traditional functional brain networks, however, are noisy and unaware of downstream prediction tasks, while also incompatible with the deep graph neural network (GNN) models. In order to fully unleash the power of GNNs in network-based fMRI analysis, we develop FBNETGEN, a task-aware and interpretable fMRI analysis framework via deep brain network generation. In particular, we formulate (1) prominent region of interest (ROI) features extraction, (2) brain networks generation, and (3) clinical predictions with GNNs, in an end-to-end trainable model under the guidance of particular prediction tasks. Along with the process, the key novel component is the graph generator which learns to transform raw time-series features into task-oriented brain networks. Our learnable graphs also provide unique interpretations by highlighting prediction-related brain regions. Comprehensive experiments on two datasets, i.e., the recently released and currently largest publicly available fMRI dataset Adolescent Brain Cognitive Development (ABCD), and the widely-used fMRI dataset PNC, prove the superior effectiveness and interpretability of FBNETGEN. The implementation is available at https://github.com/Wayfear/FBNETGEN.

Ran Xu, Hejie Cui, Yue Yu et al. · gatech

Clinical natural language processing requires methods that can address domain-specific challenges, such as complex medical terminology and clinical contexts. Recently, large language models (LLMs) have shown promise in this domain. Yet, their direct deployment can lead to privacy issues and are constrained by resources. To address this challenge, we delve into synthetic clinical text generation using LLMs for clinical NLP tasks. We propose an innovative, resource-efficient approach, ClinGen, which infuses knowledge into the process. Our model involves clinical knowledge extraction and context-informed LLM prompting. Both clinical topics and writing styles are drawn from external domain-specific knowledge graphs and LLMs to guide data generation. Our extensive empirical study across 7 clinical NLP tasks and 16 datasets reveals that ClinGen consistently enhances performance across various tasks, effectively aligning the distribution of real datasets and significantly enriching the diversity of generated training instances. Our code is available at \url{https://github.com/ritaranx/ClinGen}.

Hejie Cui, Wei Dai, Yanqiao Zhu et al.

Human brains lie at the core of complex neurobiological systems, where the neurons, circuits, and subsystems interact in enigmatic ways. Understanding the structural and functional mechanisms of the brain has long been an intriguing pursuit for neuroscience research and clinical disorder therapy. Mapping the connections of the human brain as a network is one of the most pervasive paradigms in neuroscience. Graph Neural Networks (GNNs) have recently emerged as a potential method for modeling complex network data. Deep models, on the other hand, have low interpretability, which prevents their usage in decision-critical contexts like healthcare. To bridge this gap, we propose an interpretable framework to analyze disorder-specific Regions of Interest (ROIs) and prominent connections. The proposed framework consists of two modules: a brain-network-oriented backbone model for disease prediction and a globally shared explanation generator that highlights disorder-specific biomarkers including salient ROIs and important connections. We conduct experiments on three real-world datasets of brain disorders. The results verify that our framework can obtain outstanding performance and also identify meaningful biomarkers. All code for this work is available at https://github.com/HennyJie/IBGNN.git.

Ran Xu, Yue Yu, Hejie Cui et al.

Training deep neural networks (DNNs) with limited supervision has been a popular research topic as it can significantly alleviate the annotation burden. Self-training has been successfully applied in semi-supervised learning tasks, but one drawback of self-training is that it is vulnerable to the label noise from incorrect pseudo labels. Inspired by the fact that samples with similar labels tend to share similar representations, we develop a neighborhood-based sample selection approach to tackle the issue of noisy pseudo labels. We further stabilize self-training via aggregating the predictions from different rounds during sample selection. Experiments on eight tasks show that our proposed method outperforms the strongest self-training baseline with 1.83% and 2.51% performance gain for text and graph datasets on average. Our further analysis demonstrates that our proposed data selection strategy reduces the noise of pseudo labels by 36.8% and saves 57.3% of the time when compared with the best baseline. Our code and appendices will be uploaded to https://github.com/ritaranx/NeST.

Hejie Cui, Wei Dai, Yanqiao Zhu et al.

Mapping the connectome of the human brain using structural or functional connectivity has become one of the most pervasive paradigms for neuroimaging analysis. Recently, Graph Neural Networks (GNNs) motivated from geometric deep learning have attracted broad interest due to their established power for modeling complex networked data. Despite their superior performance in many fields, there has not yet been a systematic study of how to design effective GNNs for brain network analysis. To bridge this gap, we present BrainGB, a benchmark for brain network analysis with GNNs. BrainGB standardizes the process by (1) summarizing brain network construction pipelines for both functional and structural neuroimaging modalities and (2) modularizing the implementation of GNN designs. We conduct extensive experiments on datasets across cohorts and modalities and recommend a set of general recipes for effective GNN designs on brain networks. To support open and reproducible research on GNN-based brain network analysis, we host the BrainGB website at https://braingb.us with models, tutorials, examples, as well as an out-of-box Python package. We hope that this work will provide useful empirical evidence and offer insights for future research in this novel and promising direction.

Yuwei Zhang, Chengyu Dong, Shuowei Jin et al.

Context management enables agentic models to solve long-horizon tasks through iterative summarization of previous interaction histories. However, this process typically incurs substantial decoding overhead for the extra summarization tokens, which significantly affect the end-to-end response latency at deployment. In this paper, we introduce CoMem, a novel framework that decouples memory management from the primary agent workflow, enabling these processes to execute in parallel. We propose a $k$-step-off asynchronous pipeline that overlaps the memory model's summarization with the agent's inference, effectively masking the latency of context processing. To ensure robustness under this asynchronous setting, we introduce a reward-driven training strategy that aligns the memory model to capture sufficient statistics for the agent's decision-making. Theoretical analysis confirms that CoMem offers a superior efficiency-effectiveness trade-off compared to coupled architectures. Our extensive experimental results on SWE-Bench-Verified show that CoMem provides 1.4x latency improvements upon vanilla long-context solutions while preserving most of the performance. Furthermore, we demonstrate that these latency gains scale favorably with increased system throughput, offering a modular path forward for the independent optimization of agent reasoning and memory compression.

Yi Yang, Yanqiao Zhu, Hejie Cui et al.

Brain networks characterize complex connectivities among brain regions as graph structures, which provide a powerful means to study brain connectomes. In recent years, graph neural networks have emerged as a prevalent paradigm of learning with structured data. However, most brain network datasets are limited in sample sizes due to the relatively high cost of data acquisition, which hinders the deep learning models from sufficient training. Inspired by meta-learning that learns new concepts fast with limited training examples, this paper studies data-efficient training strategies for analyzing brain connectomes in a cross-dataset setting. Specifically, we propose to meta-train the model on datasets of large sample sizes and transfer the knowledge to small datasets. In addition, we also explore two brain-network-oriented designs, including atlas transformation and adaptive task reweighing. Compared to other pre-training strategies, our meta-learning-based approach achieves higher and stabler performance, which demonstrates the effectiveness of our proposed solutions. The framework is also able to derive new insights regarding the similarities among datasets and diseases in a data-driven fashion.

Hejie Cui, Jiaying Lu, Ran Xu et al.

This comprehensive review aims to provide an overview of the current state of Healthcare Knowledge Graphs (HKGs), including their construction, utilization models, and applications across various healthcare and biomedical research domains. We thoroughly analyzed existing literature on HKGs, covering their construction methodologies, utilization techniques, and applications in basic science research, pharmaceutical research and development, clinical decision support, and public health. The review encompasses both model-free and model-based utilization approaches and the integration of HKGs with large language models (LLMs). We searched Google Scholar for relevant papers on HKGs and classified them into the following topics: HKG construction, HKG utilization, and their downstream applications in various domains. We also discussed their special challenges and the promise for future work. The review highlights the potential of HKGs to significantly impact biomedical research and clinical practice by integrating vast amounts of biomedical knowledge from multiple domains. The synergy between HKGs and LLMs offers promising opportunities for constructing more comprehensive knowledge graphs and improving the accuracy of healthcare applications. HKGs have emerged as a powerful tool for structuring medical knowledge, with broad applications across biomedical research, clinical decision-making, and public health. This survey serves as a roadmap for future research and development in the field of HKGs, highlighting the potential of combining knowledge graphs with advanced machine learning models for healthcare transformation.

Weiqi Wang, Xin Liu, Binxuan Huang et al.

RLVR is now a standard way to train LLMs on reasoning tasks with verifiable outcomes, but when rollout generation dominates the cost, efficiency depends heavily on which prompts you sample and when. In practice, prompt pools are often static or only loosely tied to the model's learning progress, so uniform sampling can't keep up with the shifting capability frontier and ends up wasting rollouts on prompts that are already solved or still out of reach. Existing approaches improve efficiency through filtering, curricula, adaptive rollout allocation, or teacher guidance, but they typically assume a fixed pool-which makes it hard to support stable on-policy pool growth-or they add extra teacher cost and latency. We introduce HeaPA (Heap Sampling and On-Policy Query Augmentation), which maintains a bounded, evolving pool, tracks the frontier using heap-based boundary sampling, expands the pool via on-policy augmentation with lightweight asynchronous validation, and stabilizes correlated queries through topology-aware re-estimation of pool statistics and controlled reinsertion. Across two training corpora, two training recipes, and seven benchmarks, HeaPA consistently improves accuracy and reaches target performance with fewer computations while keeping wall-clock time comparable. Our analyses suggest these gains come from frontier-focused sampling and on-policy pool growth, with the benefits becoming larger as model scale increases. Our code is available at https://github.com/horizon-rl/HeaPA.

Xuan Kan, Zimu Li, Hejie Cui et al.

Biological networks are commonly used in biomedical and healthcare domains to effectively model the structure of complex biological systems with interactions linking biological entities. However, due to their characteristics of high dimensionality and low sample size, directly applying deep learning models on biological networks usually faces severe overfitting. In this work, we propose R-MIXUP, a Mixup-based data augmentation technique that suits the symmetric positive definite (SPD) property of adjacency matrices from biological networks with optimized training efficiency. The interpolation process in R-MIXUP leverages the log-Euclidean distance metrics from the Riemannian manifold, effectively addressing the swelling effect and arbitrarily incorrect label issues of vanilla Mixup. We demonstrate the effectiveness of R-MIXUP with five real-world biological network datasets on both regression and classification tasks. Besides, we derive a commonly ignored necessary condition for identifying the SPD matrices of biological networks and empirically study its influence on the model performance. The code implementation can be found in Appendix E.

Hejie Cui, Rongmei Lin, Nasser Zalmout et al.

Information extraction, e.g., attribute value extraction, has been extensively studied and formulated based only on text. However, many attributes can benefit from image-based extraction, like color, shape, pattern, among others. The visual modality has long been underutilized, mainly due to multimodal annotation difficulty. In this paper, we aim to patch the visual modality to the textual-established attribute information extractor. The cross-modality integration faces several unique challenges: (C1) images and textual descriptions are loosely paired intra-sample and inter-samples; (C2) images usually contain rich backgrounds that can mislead the prediction; (C3) weakly supervised labels from textual-established extractors are biased for multimodal training. We present PV2TEA, an encoder-decoder architecture equipped with three bias reduction schemes: (S1) Augmented label-smoothed contrast to improve the cross-modality alignment for loosely-paired image and text; (S2) Attention-pruning that adaptively distinguishes the visual foreground; (S3) Two-level neighborhood regularization that mitigates the label textual bias via reliability estimation. Empirical results on real-world e-Commerce datasets demonstrate up to 11.74% absolute (20.97% relatively) F1 increase over unimodal baselines.

Xuan Kan, Antonio Aodong Chen Gu, Hejie Cui et al.

Recent neuroimaging studies have highlighted the importance of network-centric brain analysis, particularly with functional magnetic resonance imaging. The emergence of Deep Neural Networks has fostered a substantial interest in predicting clinical outcomes and categorizing individuals based on brain networks. However, the conventional approach involving static brain network analysis offers limited potential in capturing the dynamism of brain function. Although recent studies have attempted to harness dynamic brain networks, their high dimensionality and complexity present substantial challenges. This paper proposes a novel methodology, Dynamic bRAin Transformer (DART), which combines static and dynamic brain networks for more effective and nuanced brain function analysis. Our model uses the static brain network as a baseline, integrating dynamic brain networks to enhance performance against traditional methods. We innovatively employ attention mechanisms, enhancing model explainability and exploiting the dynamic brain network's temporal variations. The proposed approach offers a robust solution to the low signal-to-noise ratio of blood-oxygen-level-dependent signals, a recurring issue in direct DNN modeling. It also provides valuable insights into which brain circuits or dynamic networks contribute more to final predictions. As such, DRAT shows a promising direction in neuroimaging studies, contributing to the comprehensive understanding of brain organization and the role of neural circuits.

Hejie Cui, Xinyu Fang, Zihan Zhang et al.

Images contain rich relational knowledge that can help machines understand the world. Existing methods on visual knowledge extraction often rely on the pre-defined format (e.g., sub-verb-obj tuples) or vocabulary (e.g., relation types), restricting the expressiveness of the extracted knowledge. In this work, we take a first exploration to a new paradigm of open visual knowledge extraction. To achieve this, we present OpenVik which consists of an open relational region detector to detect regions potentially containing relational knowledge and a visual knowledge generator that generates format-free knowledge by prompting the large multimodality model with the detected region of interest. We also explore two data enhancement techniques for diversifying the generated format-free visual knowledge. Extensive knowledge quality evaluations highlight the correctness and uniqueness of the extracted open visual knowledge by OpenVik. Moreover, integrating our extracted knowledge across various visual reasoning applications shows consistent improvements, indicating the real-world applicability of OpenVik.

Haixin Wang, Hejie Cui, Chenwei Zhang et al.

Recent progress in multi-turn reinforcement learning (RL) has significantly improved reasoning LLMs' performances on complex interactive tasks. Despite advances in stabilization techniques such as fine-grained credit assignment and trajectory filtering, instability remains pervasive and often leads to training collapse. We argue that this instability stems from inefficient exploration in multi-turn settings, where policies continue to generate low-information actions that neither reduce uncertainty nor advance task progress. To address this issue, we propose Token- and Turn-level Policy Optimization (T$^2$PO), an uncertainty-aware framework that explicitly controls exploration at fine-grained levels. At the token level, T$^2$PO monitors uncertainty dynamics and triggers a thinking intervention once the marginal uncertainty change falls below a threshold. At the turn level, T$^2$PO identifies interactions with negligible exploration progress and dynamically resamples such turns to avoid wasted rollouts. We evaluate T$^2$PO in diverse environments, including WebShop, ALFWorld, and Search QA, demonstrating substantial gains in training stability and performance improvements with better exploration efficiency. Code is available at: https://github.com/WillDreamer/T2PO.

Wei Dai, Peilin Chen, Malinda Lu et al.

Recent advances in clinical AI have enabled remarkable progress across many clinical domains. However, existing benchmarks and models are primarily limited to a small set of modalities and tasks, which hinders the development of large-scale multimodal methods that can make holistic assessments of patient health and well-being. To bridge this gap, we introduce Clinical Large-Scale Integrative Multimodal Benchmark (CLIMB), a comprehensive clinical benchmark unifying diverse clinical data across imaging, language, temporal, and graph modalities. CLIMB comprises 4.51 million patient samples totaling 19.01 terabytes distributed across 2D imaging, 3D video, time series, graphs, and multimodal data. Through extensive empirical evaluation, we demonstrate that multitask pretraining significantly improves performance on understudied domains, achieving up to 29% improvement in ultrasound and 23% in ECG analysis over single-task learning. Pretraining on CLIMB also effectively improves models' generalization capability to new tasks, and strong unimodal encoder performance translates well to multimodal performance when paired with task-appropriate fusion strategies. Our findings provide a foundation for new architecture designs and pretraining strategies to advance clinical AI research. Code is released at https://github.com/DDVD233/climb.

Asad Aali, Muhammad Ahmed Mohsin, Vasiliki Bikia et al.

As language models (LMs) are increasingly adopted across domains, high-quality benchmarking frameworks that accurately estimate performance are essential for guiding deployment decisions. While frameworks such as Holistic Evaluation of Language Models (HELM) enable broad evaluation across tasks, they often rely on fixed prompts that fail to generalize across LMs, yielding unrepresentative performance estimates. Unless we approximate each LM's ceiling (maximum achievable via changes to the prompt), we risk underestimating performance. Declarative prompting frameworks, such as DSPy, offer a scalable alternative to manual prompt engineering by crafting structured prompts that can be optimized per task. However, such frameworks have not been systematically evaluated across established benchmarks. We present a reproducible DSPy+HELM framework that introduces structured prompting methods which elicit reasoning, enabling more accurate LM benchmarking. Using four prompting methods, we evaluate four frontier LMs across seven benchmarks (general/medical domain) against existing HELM baseline scores. We find that without structured prompting: (i) HELM underestimates LM performance (by 4% average), (ii) performance estimates vary more across benchmarks ($+$2% standard deviation), (iii) performance gaps are misrepresented (leaderboard rankings flip on 3/7 benchmarks), and (iv) introducing chain-of-thought reduces LM sensitivity to prompt design (smaller $Δ$ across prompts). To our knowledge, this is the first benchmarking study to systematically integrate structured prompting into an established evaluation framework, demonstrating how scalable performance-ceiling approximation yields more robust, decision-useful benchmarks. We open-source (i) DSPy+HELM Integration (https://github.com/stanford-crfm/helm/pull/3893) and (ii) Prompt Optimization Pipeline (https://github.com/StanfordMIMI/dspy-helm).

Ziyang Zhang, Hejie Cui, Ran Xu et al.

The growing availability of well-organized Electronic Health Records (EHR) data has enabled the development of various machine learning models towards disease risk prediction. However, existing risk prediction methods overlook the heterogeneity of complex diseases, failing to model the potential disease subtypes regarding their corresponding patient visits and clinical concept subgroups. In this work, we introduce TACCO, a novel framework that jointly discovers clusters of clinical concepts and patient visits based on a hypergraph modeling of EHR data. Specifically, we develop a novel self-supervised co-clustering framework that can be guided by the risk prediction task of specific diseases. Furthermore, we enhance the hypergraph model of EHR data with textual embeddings and enforce the alignment between the clusters of clinical concepts and patient visits through a contrastive objective. Comprehensive experiments conducted on the public MIMIC-III dataset and Emory internal CRADLE dataset over the downstream clinical tasks of phenotype classification and cardiovascular risk prediction demonstrate an average 31.25% performance improvement compared to traditional ML baselines and a 5.26% improvement on top of the vanilla hypergraph model without our co-clustering mechanism. In-depth model analysis, clustering results analysis, and clinical case studies further validate the improved utilities and insightful interpretations delivered by TACCO. Code is available at https://github.com/PericlesHat/TACCO.

Wei Dai, Hejie Cui, Xuan Kan et al.

Brain networks, graphical models such as those constructed from MRI, have been widely used in pathological prediction and analysis of brain functions. Within the complex brain system, differences in neuronal connection strengths parcellate the brain into various functional modules (network communities), which are critical for brain analysis. However, identifying such communities within the brain has been a nontrivial issue due to the complexity of neuronal interactions. In this work, we propose a novel interpretable transformer-based model for joint hierarchical cluster identification and brain network classification. Extensive experimental results on real-world brain network datasets show that with the help of hierarchical clustering, the model achieves increased accuracy and reduced runtime complexity while providing plausible insight into the functional organization of brain regions. The implementation is available at https://github.com/DDVD233/THC.

Hejie Cui, Jiaying Lu, Yao Ge et al.

Graph neural networks (GNNs), as a group of powerful tools for representation learning on irregular data, have manifested superiority in various downstream tasks. With unstructured texts represented as concept maps, GNNs can be exploited for tasks like document retrieval. Intrigued by how can GNNs help document retrieval, we conduct an empirical study on a large-scale multi-discipline dataset CORD-19. Results show that instead of the complex structure-oriented GNNs such as GINs and GATs, our proposed semantics-oriented graph functions achieve better and more stable performance based on the BM25 retrieved candidates. Our insights in this case study can serve as a guideline for future work to develop effective GNNs with appropriate semantics-oriented inductive biases for textual reasoning tasks like document retrieval and classification. All code for this case study is available at https://github.com/HennyJie/GNN-DocRetrieval.

Hejie Cui, Zijie Lu, Pan Li et al.

Graph neural networks (GNNs) have been widely used in various graph-related problems such as node classification and graph classification, where superior performance is mainly established when natural node features are available. However, it is not well understood how GNNs work without natural node features, especially regarding the various ways to construct artificial ones. In this paper, we point out the two types of artificial node features, i.e., positional and structural node features, and provide insights on why each of them is more appropriate for certain tasks, i.e., positional node classification, structural node classification, and graph classification. Extensive experimental results on 10 benchmark datasets validate our insights, thus leading to a practical guideline on the choices between different artificial node features for GNNs on non-attributed graphs. The code is available at https://github.com/zjzijielu/gnn-positional-structural-node-features.

Hejie Cui, Xinyu Fang, Ran Xu et al.

Electronic Health Records (EHRs) have become increasingly popular to support clinical decision-making and healthcare in recent decades. EHRs usually contain heterogeneous information, such as structural data in tabular form and unstructured data in textual notes. Different types of information in EHRs can complement each other and provide a more complete picture of the health status of a patient. While there has been a lot of research on representation learning of structured EHR data, the fusion of different types of EHR data (multimodal fusion) is not well studied. This is mostly because of the complex medical coding systems used and the noise and redundancy present in the written notes. In this work, we propose a new framework called MINGLE, which integrates both structures and semantics in EHR effectively. Our framework uses a two-level infusion strategy to combine medical concept semantics and clinical note semantics into hypergraph neural networks, which learn the complex interactions between different types of data to generate visit representations for downstream prediction. Experiment results on two EHR datasets, the public MIMIC-III and private CRADLE, show that MINGLE can effectively improve predictive performance by 11.83% relatively, enhancing semantic integration as well as multimodal fusion for structural and textual EHR data.

Hejie Cui, Alyssa Unell, Bowen Chen et al. · stanford

Large language models (LLMs) have emerged as promising tools for assisting in medical tasks, yet processing Electronic Health Records (EHRs) presents unique challenges due to their longitudinal nature. While LLMs' capabilities to perform medical tasks continue to improve, their ability to reason over temporal dependencies across multiple patient visits and time frames remains unexplored. We introduce TIMER (Temporal Instruction Modeling and Evaluation for Longitudinal Clinical Records), a framework that incorporate instruction-response pairs grounding to different parts of a patient's record as a critical dimension in both instruction evaluation and tuning for longitudinal clinical records. We develop TIMER-Bench, the first time-aware benchmark that evaluates temporal reasoning capabilities over longitudinal EHRs, as well as TIMER-Instruct, an instruction-tuning methodology for LLMs to learn reasoning over time. We demonstrate that models fine-tuned with TIMER-Instruct improve performance by 7.3% on human-generated benchmarks and 9.2% on TIMER-Bench, indicating that temporal instruction-tuning improves model performance for reasoning over EHR.

Yuzhang Xie, Hejie Cui, Ziyang Zhang et al.

Medical diagnosis prediction plays a critical role in disease detection and personalized healthcare. While machine learning (ML) models have been widely adopted for this task, their reliance on supervised training limits their ability to generalize to unseen cases, particularly given the high cost of acquiring large, labeled datasets. Large language models (LLMs) have shown promise in leveraging language abilities and biomedical knowledge for diagnosis prediction. However, they often suffer from hallucinations, lack structured medical reasoning, and produce useless outputs. To address these challenges, we propose KERAP, a knowledge graph (KG)-enhanced reasoning approach that improves LLM-based diagnosis prediction through a multi-agent architecture. Our framework consists of a linkage agent for attribute mapping, a retrieval agent for structured knowledge extraction, and a prediction agent that iteratively refines diagnosis predictions. Experimental results demonstrate that KERAP enhances diagnostic reliability efficiently, offering a scalable and interpretable solution for zero-shot medical diagnosis prediction.

Yanbang Wang, Hejie Cui, Jon Kleinberg

Graphs data is crucial for many applications, and much of it exists in the relations described in textual format. As a result, being able to accurately recall and encode a graph described in earlier text is a basic yet pivotal ability that LLMs need to demonstrate if they are to perform reasoning tasks that involve graph-structured information. Human performance at graph recall has been studied by cognitive scientists for decades, and has been found to often exhibit certain structural patterns of bias that align with human handling of social relationships. To date, however, we know little about how LLMs behave in analogous graph recall tasks: do their recalled graphs also exhibit certain biased patterns, and if so, how do they compare with humans and affect other graph reasoning tasks? In this work, we perform the first systematical study of graph recall by LLMs, investigating the accuracy and biased microstructures (local structural patterns) in their recall. We find that LLMs not only underperform often in graph recall, but also tend to favor more triangles and alternating 2-paths. Moreover, we find that more advanced LLMs have a striking dependence on the domain that a real-world graph comes from -- by yielding the best recall accuracy when the graph is narrated in a language style consistent with its original domain.

Amin Adibi, Xu Cao, Zongliang Ji et al.

The fourth Machine Learning for Health (ML4H) symposium was held in person on December 15th and 16th, 2024, in the traditional, ancestral, and unceded territories of the Musqueam, Squamish, and Tsleil-Waututh Nations in Vancouver, British Columbia, Canada. The symposium included research roundtable sessions to foster discussions between participants and senior researchers on timely and relevant topics for the ML4H community. The organization of the research roundtables at the conference involved 13 senior and 27 junior chairs across 13 tables. Each roundtable session included an invited senior chair (with substantial experience in the field), junior chairs (responsible for facilitating the discussion), and attendees from diverse backgrounds with an interest in the session's topic.

Genglin Liu, Shijie Geng, Sha Li et al.

Multimodal LLM-powered agents have recently demonstrated impressive capabilities in web navigation, enabling agents to complete complex browsing tasks across diverse domains. However, current agents struggle with repetitive errors and lack the ability to learn from past experiences across sessions, limiting their long-term robustness and sample efficiency. We introduce WebCoach, a model-agnostic self-evolving framework that equips web browsing agents with persistent cross-session memory, enabling improved long-term planning, reflection, and continual learning without retraining. WebCoach consists of three key components: (1) a WebCondenser, which standardizes raw navigation logs into concise summaries; (2) an External Memory Store, which organizes complete trajectories as episodic experiences; and (3) a Coach, which retrieves relevant experiences based on similarity and recency, and decides whether to inject task-specific advice into the agent via runtime hooks. This design empowers web agents to access long-term memory beyond their native context window, improving robustness in complex browsing tasks. Moreover, WebCoach achieves self-evolution by continuously curating episodic memory from new navigation trajectories, enabling agents to improve over time without retraining. Evaluations on the WebVoyager benchmark demonstrate that WebCoach consistently improves the performance of browser-use agents across three different LLM backbones. With a 38B model, it increases task success rates from 47% to 61% while reducing or maintaining the average number of steps. Notably, smaller base models with WebCoach achieve performance comparable to the same web agent using GPT-4o.

Hejie Cui, Lingjun Mao, Xin Liang et al.

Recent advancements in multimodal foundation models have showcased impressive capabilities in understanding and reasoning with visual and textual information. Adapting these foundation models trained for general usage to specialized domains like biomedicine requires large-scale domain-specific instruction datasets. While existing works have explored curating such datasets automatically, the resultant datasets are not explicitly aligned with domain expertise. In this work, we propose a data-centric framework, Biomedical Visual Instruction Tuning with Clinician Preference Alignment (BioMed-VITAL), that incorporates clinician preferences into both stages of generating and selecting instruction data for tuning biomedical multimodal foundation models. First, during the generation stage, we prompt the GPT-4V generator with a diverse set of clinician-selected demonstrations for preference-aligned data candidate generation. Then, during the selection phase, we train a separate selection model, which explicitly distills clinician and policy-guided model preferences into a rating function to select high-quality data for medical instruction tuning. Results show that the model tuned with the instruction-following data from our method demonstrates a significant improvement in open visual chat (18.5% relatively) and medical VQA (win rate up to 81.73%). Our instruction-following data and models are available at BioMed-VITAL.github.io.

Hejie Cui, Zhuocheng Shen, Jieyu Zhang et al.

Electronic health records (EHRs) contain valuable patient data for health-related prediction tasks, such as disease prediction. Traditional approaches rely on supervised learning methods that require large labeled datasets, which can be expensive and challenging to obtain. In this study, we investigate the feasibility of applying Large Language Models (LLMs) to convert structured patient visit data (e.g., diagnoses, labs, prescriptions) into natural language narratives. We evaluate the zero-shot and few-shot performance of LLMs using various EHR-prediction-oriented prompting strategies. Furthermore, we propose a novel approach that utilizes LLM agents with different roles: a predictor agent that makes predictions and generates reasoning processes and a critic agent that analyzes incorrect predictions and provides guidance for improving the reasoning of the predictor agent. Our results demonstrate that with the proposed approach, LLMs can achieve decent few-shot performance compared to traditional supervised learning methods in EHR-based disease predictions, suggesting its potential for health-oriented applications.

Chen Ling, Xujiang Zhao, Jiaying Lu et al.

Large language models (LLMs) have significantly advanced the field of natural language processing (NLP), providing a highly useful, task-agnostic foundation for a wide range of applications. However, directly applying LLMs to solve sophisticated problems in specific domains meets many hurdles, caused by the heterogeneity of domain data, the sophistication of domain knowledge, the uniqueness of domain objectives, and the diversity of the constraints (e.g., various social norms, cultural conformity, religious beliefs, and ethical standards in the domain applications). Domain specification techniques are key to make large language models disruptive in many applications. Specifically, to solve these hurdles, there has been a notable increase in research and practices conducted in recent years on the domain specialization of LLMs. This emerging field of study, with its substantial potential for impact, necessitates a comprehensive and systematic review to better summarize and guide ongoing work in this area. In this article, we present a comprehensive survey on domain specification techniques for large language models, an emerging direction critical for large language model applications. First, we propose a systematic taxonomy that categorizes the LLM domain-specialization techniques based on the accessibility to LLMs and summarizes the framework for all the subcategories as well as their relations and differences to each other. Second, we present an extensive taxonomy of critical application domains that can benefit dramatically from specialized LLMs, discussing their practical significance and open challenges. Last, we offer our insights into the current research status and future trends in this area.

Yi Yang, Hejie Cui, Carl Yang

The human brain is the central hub of the neurobiological system, controlling behavior and cognition in complex ways. Recent advances in neuroscience and neuroimaging analysis have shown a growing interest in the interactions between brain regions of interest (ROIs) and their impact on neural development and disorder diagnosis. As a powerful deep model for analyzing graph-structured data, Graph Neural Networks (GNNs) have been applied for brain network analysis. However, training deep models requires large amounts of labeled data, which is often scarce in brain network datasets due to the complexities of data acquisition and sharing restrictions. To make the most out of available training data, we propose PTGB, a GNN pre-training framework that captures intrinsic brain network structures, regardless of clinical outcomes, and is easily adaptable to various downstream tasks. PTGB comprises two key components: (1) an unsupervised pre-training technique designed specifically for brain networks, which enables learning from large-scale datasets without task-specific labels; (2) a data-driven parcellation atlas mapping pipeline that facilitates knowledge transfer across datasets with different ROI systems. Extensive evaluations using various GNN models have demonstrated the robust and superior performance of PTGB compared to baseline methods.

Yanqiao Zhu, Yichen Xu, Hejie Cui et al.

Recently, heterogeneous Graph Neural Networks (GNNs) have become a de facto model for analyzing HGs, while most of them rely on a relative large number of labeled data. In this work, we investigate Contrastive Learning (CL), a key component in self-supervised approaches, on HGs to alleviate the label scarcity problem. We first generate multiple semantic views according to metapaths and network schemas. Then, by pushing node embeddings corresponding to different semantic views close to each other (positives) and pulling other embeddings apart (negatives), one can obtain informative representations without human annotations. However, this CL approach ignores the relative hardness of negative samples, which may lead to suboptimal performance. Considering the complex graph structure and the smoothing nature of GNNs, we propose a structure-aware hard negative mining scheme that measures hardness by structural characteristics for HGs. By synthesizing more negative nodes, we give larger weights to harder negatives with limited computational overhead to further boost the performance. Empirical studies on three real-world datasets show the effectiveness of our proposed method. The proposed method consistently outperforms existing state-of-the-art methods and notably, even surpasses several supervised counterparts.

Xuan Kan, Hejie Cui, Ying Guo et al.

Recent studies in neuroscience show great potential of functional brain networks constructed from fMRI data for popularity modeling and clinical predictions. However, existing functional brain networks are noisy and unaware of downstream prediction tasks, while also incompatible with recent powerful machine learning models of GNNs. In this work, we develop an end-to-end trainable pipeline to extract prominent fMRI features, generate brain networks, and make predictions with GNNs, all under the guidance of downstream prediction tasks. Preliminary experiments on the PNC fMRI data show the superior effectiveness and unique interpretability of our framework.

Hejie Cui, Wei Dai, Yanqiao Zhu et al.

Interpretable brain network models for disease prediction are of great value for the advancement of neuroscience. GNNs are promising to model complicated network data, but they are prone to overfitting and suffer from poor interpretability, which prevents their usage in decision-critical scenarios like healthcare. To bridge this gap, we propose BrainNNExplainer, an interpretable GNN framework for brain network analysis. It is mainly composed of two jointly learned modules: a backbone prediction model that is specifically designed for brain networks and an explanation generator that highlights disease-specific prominent brain network connections. Extensive experimental results with visualizations on two challenging disease prediction datasets demonstrate the unique interpretability and outstanding performance of BrainNNExplainer.

Xuan Kan, Hejie Cui, Carl Yang

Relation prediction among entities in images is an important step in scene graph generation (SGG), which further impacts various visual understanding and reasoning tasks. Existing SGG frameworks, however, require heavy training yet are incapable of modeling unseen (i.e.,zero-shot) triplets. In this work, we stress that such incapability is due to the lack of commonsense reasoning,i.e., the ability to associate similar entities and infer similar relations based on general understanding of the world. To fill this gap, we propose CommOnsense-integrAted sCenegrapHrElation pRediction (COACHER), a framework to integrate commonsense knowledge for SGG, especially for zero-shot relation prediction. Specifically, we develop novel graph mining pipelines to model the neighborhoods and paths around entities in an external commonsense knowledge graph, and integrate them on top of state-of-the-art SGG frameworks. Extensive quantitative evaluations and qualitative case studies on both original and manipulated datasets from Visual Genome demonstrate the effectiveness of our proposed approach.

Yanqiao Zhu, Hejie Cui, Lifang He et al.

Multimodal brain networks characterize complex connectivities among different brain regions from both structural and functional aspects and provide a new means for mental disease analysis. Recently, Graph Neural Networks (GNNs) have become a de facto model for analyzing graph-structured data. However, how to employ GNNs to extract effective representations from brain networks in multiple modalities remains rarely explored. Moreover, as brain networks provide no initial node features, how to design informative node attributes and leverage edge weights for GNNs to learn is left unsolved. To this end, we develop a novel multiview GNN for multimodal brain networks. In particular, we regard each modality as a view for brain networks and employ contrastive learning for multimodal fusion. Then, we propose a GNN model which takes advantage of the message passing scheme by propagating messages based on degree statistics and brain region connectivities. Extensive experiments on two real-world disease datasets (HIV and Bipolar) demonstrate the effectiveness of our proposed method over state-of-the-art baselines.

Hejie Cui, Xinglong Liu, Ning Huang

Pulmonary vessel segmentation is important for clinical diagnosis of pulmonary diseases, while is also challenging due to the complicated structure. In this work, we present an effective framework and refinement process of pulmonary vessel segmentation from chest computed tomographic (CT) images. The key to our approach is a 2.5D segmentation network applied from three orthogonal axes, which presents a robust and fully automated pulmonary vessel segmentation result with lower network complexity and memory usage compared to 3D networks. The slice radius is introduced to convolve the adjacent information of the center slice and the multi-planar fusion optimizes the presentation of intra- and inter- slice features. Besides, the tree-like structure of the pulmonary vessel is extracted in the post-processing process, which is used for segmentation refining and pruning. In the evaluation experiments, three fusion methods are tested and the most promising one is compared with the state-of-the-art 2D and 3D structures on 300 cases of lung images randomly selected from LIDC dataset. Our method outperforms other network structures by a large margin and achieves by far the highest average DICE score of 0.9272 and precision of 0.9310, as per our knowledge from the pulmonary vessel segmentation models available in the literature.