Searcharxiv⌕ Search

arXiv subjects

John Wu

Publications and source records attributed to John Wu.

16 recordsLinked to original sources
arXiv2026

A Practical Guide Towards Interpreting Time-Series Deep Clinical Predictive Models: A Reproducibility Study

Clinical decisions are high-stakes and require explicit justification, making model interpretability essential for auditing deep clinical models prior to deployment. As the ecosystem of model architectures and explainability methods expands, critical questions remain: Do architectural features like attention improve explainability? Do interpretability approaches generalize across clinical tasks? While prior benchmarking efforts exist, they often lack extensibility and reproducibility, and critically, fail to systematically examine how interpretability varies across the interplay of clinical tasks and model architectures. To address these gaps, we present a comprehensive benchmark evaluating interpretability methods across diverse clinical prediction tasks and model architectures. Our analysis reveals that: (1) attention when leveraged properly is a highly efficient approach for faithfully interpreting model predictions; (2) black-box interpreters like KernelSHAP and LIME are computationally infeasible for time-series clinical prediction tasks; and (3) several interpretability approaches are too unreliable to be trustworthy. From our findings, we discuss several guidelines on improving interpretability within clinical predictive pipelines. To support reproducibility and extensibility, we provide our implementations via PyHealth, a well-documented open-source framework: https://github.com/sunlabuiuc/PyHealth.

cs.LG↗
arXiv2026

A Modular LLM Framework for Explainable Price Outlier Detection

Detecting product price outliers is important for retail and e-commerce stores as erroneous or unexpectedly high prices adversely affect competitiveness, revenue, and consumer trust. Classical techniques offer simple thresholds while ignoring the rich semantic relationships among product attributes. We propose an agentic Large Language Model (LLM) framework that treats outlier price flagging as a reasoning task grounded in related product detection and comparison. The system processes the prices of target products in three stages: (i) relevance classification selects price-relevant similar products using product descriptions and attributes; (ii) relative utility assessment evaluates the target product against each similar product along price influencing dimensions (e.g., brand, size, features); (iii) reasoning-based decision aggregates these justifications into an explainable price outlier judgment. The framework attains over 75% agreement with human auditors on a test dataset, and outperforms zero-shot and retrieval based LLM techniques. Ablation studies show the sensitivity of the method to key hyper-parameters and testify on its flexibility to be applied to cases with different accuracy requirement and auditor agreements.

cs.CL↗
arXiv2026

Bridging the Reproducibility Divide: Open Source Software's Role in Standardizing Healthcare AI

Our analysis of recent AI4H publications reveals that, despite a trend toward utilizing open datasets and sharing modeling code, 74% of AI4H papers still rely on private datasets or do not share their code. This is especially concerning in healthcare applications, where trust is essential. Furthermore, inconsistent and poorly documented data preprocessing pipelines result in variable model performance reports, even for identical tasks and datasets, making it challenging to evaluate the true effectiveness of AI models. Despite the challenges posed by the reproducibility crisis, addressing these issues through open practices offers substantial benefits. For instance, while the reproducibility mandate adds extra effort to research and publication, it significantly enhances the impact of the work. Our analysis shows that papers that used both public datasets and shared code received, on average, 110% more citations than those that do neither--more than doubling the citation count. Given the clear benefits of enhancing reproducibility, it is imperative for the AI4H community to take concrete steps to overcome existing barriers. The community should promote open science practices, establish standardized guidelines for data preprocessing, and develop robust benchmarks. Tackling these challenges through open-source development can improve reproducibility, which is essential for ensuring that AI models are safe, effective, and beneficial for patient care. This approach will help build more trustworthy AI systems that can be integrated into healthcare settings, ultimately contributing to better patient outcomes and advancing the field of medicine.

cs.CY↗
arXiv2026

Making Conformal Predictors Robust in Healthcare Settings: a Case Study on EEG Classification

Quantifying uncertainty in clinical predictions is critical for high-stakes diagnosis tasks. Conformal prediction offers a principled approach by providing prediction sets with theoretical coverage guarantees. However, in practice, patient distribution shifts violate the i.i.d. assumptions underlying standard conformal methods, leading to poor coverage in healthcare settings. In this work, we evaluate several conformal prediction approaches on EEG seizure classification, a task with known distribution shift challenges and label uncertainty. We demonstrate that personalized calibration strategies can improve coverage by over 20 percentage points while maintaining comparable prediction set sizes. Our implementation is available via PyHealth, an open-source healthcare AI framework: https://github.com/sunlabuiuc/PyHealth.

cs.LG↗
arXiv2026

PyHealth 2.0: A Comprehensive Open-Source Toolkit for Accessible and Reproducible Clinical Deep Learning

Difficulty replicating baselines, high computational costs, and required domain expertise create persistent barriers to clinical AI research. To address these challenges, we introduce PyHealth 2.0, an enhanced clinical deep learning toolkit that enables predictive modeling in as few as 7 lines of code. PyHealth 2.0 offers three key contributions: (1) a comprehensive toolkit addressing reproducibility and compatibility challenges by unifying 15+ datasets, 20+ clinical tasks, 25+ models, 5+ interpretability methods, and uncertainty quantification including conformal prediction within a single framework that supports diverse clinical data modalities - signals, imaging, and electronic health records - with translation of 5+ medical coding standards; (2) accessibility-focused design accommodating multimodal data and diverse computational resources with up to 39x faster processing and 20x lower memory usage, enabling work from 16GB laptops to production systems; and (3) an active open-source community of 400+ members lowering domain expertise barriers through extensive documentation, reproducible research contributions, and collaborations with academic health systems and industry partners, including multi-language support via RHealth. PyHealth 2.0 establishes an open-source foundation and community advancing accessible, reproducible healthcare AI. Available at pip install pyhealth.

cs.LG↗
arXiv2025

MIMIC-RD: Can LLMs differentially diagnose rare diseases in real-world clinical settings?

Despite rare diseases affecting 1 in 10 Americans, their differential diagnosis remains challenging. Due to their impressive recall abilities, large language models (LLMs) have been recently explored for differential diagnosis. Existing approaches to evaluating LLM-based rare disease diagnosis suffer from two critical limitations: they rely on idealized clinical case studies that fail to capture real-world clinical complexity, or they use ICD codes as disease labels, which significantly undercounts rare diseases since many lack direct mappings to comprehensive rare disease databases like Orphanet. To address these limitations, we explore MIMIC-RD, a rare disease differential diagnosis benchmark constructed by directly mapping clinical text entities to Orphanet. Our methodology involved an initial LLM-based mining process followed by validation from four medical annotators to confirm identified entities were genuine rare diseases. We evaluated various models on our dataset of 145 patients and found that current state-of-the-art LLMs perform poorly on rare disease differential diagnosis, highlighting the substantial gap between existing capabilities and clinical needs. From our findings, we outline several future steps towards improving differential diagnosis of rare diseases.

cs.AI↗
arXiv2025

RDMA: Cost Effective Agent-Driven Rare Disease Mining from Electronic Health Records

Rare diseases affect 1 in 10 Americans yet remain systematically underdocumented in clinical records. ICD-based systems cannot capture their breadth, over 50\% of Orphanet codes lack a direct ICD mapping and only 2.2\% of HPO codes have matching ICD codes, leaving patient populations invisible and delaying diagnosis. Mining unstructured clinical notes offers a direct path forward, but real notes are long, noisy, and abbreviation-dense, and limited annotations make fine-tuning infeasible, demanding approaches that generalize without task-specific training. We present Rare Disease Mining Agents (RDMA), an agentic framework equipping smaller quantized LLMs with tools for abbreviation resolution, implicit phenotype reasoning, and ontology grounding against Orphanet and HPO. RDMA substantially outperforms fine-tuned and RAG-based baselines across benchmarks with different data characteristics, without any task-specific training. A small quantized model achieves maximal performance, reducing inference costs by up to 10x and local hardware costs by up to 17x, enabling private deployment on standard hardware without cloud-based PHI exposure. RDMA's uncertainty-flagging mechanism further reduces expert annotation burden while preserving agreement quality, supporting scalable rare disease documentation in clinical practice. Available at https://github.com/jhnwu3/RDMA.

cs.LG↗
arXiv2025

Recent Advances, Applications and Open Challenges in Machine Learning for Health: Reflections from Research Roundtables at ML4H 2024 Symposium

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.

cs.LG↗
arXiv2024

Beyond Label Attention: Transparency in Language Models for Automated Medical Coding via Dictionary Learning

Medical coding, the translation of unstructured clinical text into standardized medical codes, is a crucial but time-consuming healthcare practice. Though large language models (LLM) could automate the coding process and improve the efficiency of such tasks, interpretability remains paramount for maintaining patient trust. Current efforts in interpretability of medical coding applications rely heavily on label attention mechanisms, which often leads to the highlighting of extraneous tokens irrelevant to the ICD code. To facilitate accurate interpretability in medical language models, this paper leverages dictionary learning that can efficiently extract sparsely activated representations from dense language model embeddings in superposition. Compared with common label attention mechanisms, our model goes beyond token-level representations by building an interpretable dictionary which enhances the mechanistic-based explanations for each ICD code prediction, even when the highlighted tokens are medically irrelevant. We show that dictionary features can steer model behavior, elucidate the hidden meanings of upwards of 90% of medically irrelevant tokens, and are human interpretable.

cs.CL↗
arXiv2024

Generative AI for Overall Mission Effectiveness at the Habitable Worlds Observatory

Here we present several use cases for using Generative AI (Gen AI) to improve systems engineering and cognitive knowledge management related to the future of astronomy from a culmination of working meetings and presentations as part of the Gen AI Task Group for the NASA Habitable Worlds Observatory (HWO) Science and Technology Architecture Review Team (START) AI/ML Working Group. Collectively, our group mission statement is "Where is the Human-in-the-loop as Gen AI systems become more powerful and autonomous?" with an emphasis on the ethical applications of Gen AI, guided by using these systems to remove drudgery from human work while simultaneously increasing opportunities for humans to experience more collective creativity and innovation. The HWO mission stands to benefit dramatically from generative models for different data types including text, time series/spectra, and image data. These cover a wide range of applications in science and engineering for HWO, including: mission development acceleration, data analysis and interpretation, enhancing imaging capabilities, anomaly detection, predictive modeling and simulation, data augmentation for machine learning, instrument calibration and optimization, public engagement and education, and assisting in mission planning. As an example, through sensitivity analysis of simulated exoplanet population science data sets of various generative model complexity, we can reverse engineer the measurement uncertainty requirements for HWO instruments to produce data that can constrain population models and thus inform HWO design requirements. This approach to HWO design is one example of a strategy that can ensure that HWO remains AI-ready. Through presenting herein a combination of visionary ideas balanced with grounded validated use case examples, we aim to support the development of a long-term strategy to keep HWO AI-ready as it moves forward.

astro-ph.IM↗
arXiv2024

DILA: Dictionary Label Attention for Mechanistic Interpretability in High-dimensional Multi-label Medical Coding Prediction

Predicting high-dimensional or extreme multilabels, such as in medical coding, requires both accuracy and interpretability. Existing works often rely on local interpretability methods, failing to provide comprehensive explanations of the overall mechanism behind each label prediction within a multilabel set. We propose a mechanistic interpretability module called DIctionary Label Attention (\method) that disentangles uninterpretable dense embeddings into a sparse embedding space, where each nonzero element (a dictionary feature) represents a globally learned medical concept. Through human evaluations, we show that our sparse embeddings are more human understandable than its dense counterparts by at least 50 percent. Our automated dictionary feature identification pipeline, leveraging large language models (LLMs), uncovers thousands of learned medical concepts by examining and summarizing the highest activating tokens for each dictionary feature. We represent the relationships between dictionary features and medical codes through a sparse interpretable matrix, enhancing the mechanistic and global understanding of the model's predictions while maintaining competitive performance and scalability without extensive human annotation.

cs.CL↗
arXiv2024

Deep Learning Cosmic Ray Transport from Density Maps of Simulated, Turbulent Gas

The coarse-grained propagation of Galactic cosmic rays (CRs) is traditionally constrained by phenomenological models of Milky Way CR propagation fit to a variety of direct and indirect observables; however, constraining the fine-grained transport of CRs along individual magnetic field lines -- for instance, diffusive vs streaming transport models -- is an unsolved challenge. Leveraging a recent training set of magnetohydrodynamic turbulent box simulations, with CRs spanning a range of transport parameters, we use convolutional neural networks (CNNs) trained solely on gas density maps to classify CR transport regimes. We find that even relatively simple CNNs can quite effectively classify density slices to corresponding CR transport parameters, distinguishing between streaming and diffusive transport, as well as magnitude of diffusivity, with class accuracies between $92\%$ and $99\%$. As we show, the transport-dependent imprints that CRs leave on the gas are not all tied to the resulting density power spectra: classification accuracies are still high even when image spectra are flattened ($85\%$ to $98\%$ accuracy), highlighting CR transport-dependent changes to turbulent phase information. We interpret our results with saliency maps and image modifications, and we discuss physical insights and future applications.

astro-ph.GA↗
arXiv2024

If LLM Is the Wizard, Then Code Is the Wand: A Survey on How Code Empowers Large Language Models to Serve as Intelligent Agents

The prominent large language models (LLMs) of today differ from past language models not only in size, but also in the fact that they are trained on a combination of natural language and formal language (code). As a medium between humans and computers, code translates high-level goals into executable steps, featuring standard syntax, logical consistency, abstraction, and modularity. In this survey, we present an overview of the various benefits of integrating code into LLMs' training data. Specifically, beyond enhancing LLMs in code generation, we observe that these unique properties of code help (i) unlock the reasoning ability of LLMs, enabling their applications to a range of more complex natural language tasks; (ii) steer LLMs to produce structured and precise intermediate steps, which can then be connected to external execution ends through function calls; and (iii) take advantage of code compilation and execution environment, which also provides diverse feedback for model improvement. In addition, we trace how these profound capabilities of LLMs, brought by code, have led to their emergence as intelligent agents (IAs) in situations where the ability to understand instructions, decompose goals, plan and execute actions, and refine from feedback are crucial to their success on downstream tasks. Finally, we present several key challenges and future directions of empowering LLMs with code.

cs.CL↗
arXiv2022

A Machine Learning Approach to Enhancing eROSITA Observations

The eROSITA X-ray telescope, launched in 2019, is predicted to observe roughly 100,000 galaxy clusters. Follow-up observations of these clusters from Chandra, for example, will be needed to resolve outstanding questions about galaxy cluster physics. Deep Chandra cluster observations are expensive and follow-up of every eROSITA cluster is infeasible, therefore, objects chosen for follow-up must be chosen with care. To address this, we have developed an algorithm for predicting longer duration, background-free observations based on mock eROSITA observations. We make use of the hydrodynamic cosmological simulation Magneticum, have simulated eROSITA instrument conditions using SIXTE, and have applied a novel convolutional neural network to output a deep Chandra-like "super observation" of each cluster in our simulation sample. Any follow-up merit assessment tool should be designed with a specific use case in mind; our model produces observations that accurately and precisely reproduce the cluster morphology, which is a critical ingredient for determining cluster dynamical state and core type. Our model will advance our understanding of galaxy clusters by improving follow-up selection and demonstrates that image-to-image deep learning algorithms are a viable method for simulating realistic follow-up observations.

astro-ph.CO↗
arXiv2022

A Case for Dataset Specific Profiling

Data-driven science is an emerging paradigm where scientific discoveries depend on the execution of computational AI models against rich, discipline-specific datasets. With modern machine learning frameworks, anyone can develop and execute computational models that reveal concepts hidden in the data that could enable scientific applications. For important and widely used datasets, computing the performance of every computational model that can run against a dataset is cost prohibitive in terms of cloud resources. Benchmarking approaches used in practice use representative datasets to infer performance without actually executing models. While practicable, these approaches limit extensive dataset profiling to a few datasets and introduce bias that favors models suited for representative datasets. As a result, each dataset's unique characteristics are left unexplored and subpar models are selected based on inference from generalized datasets. This necessitates a new paradigm that introduces dataset profiling into the model selection process. To demonstrate the need for dataset-specific profiling, we answer two questions:(1) Can scientific datasets significantly permute the rank order of computational models compared to widely used representative datasets? (2) If so, could lightweight model execution improve benchmarking accuracy? Taken together, the answers to these questions lay the foundation for a new dataset-aware benchmarking paradigm.

cs.LG↗