Searcharxiv⌕ Search

arXiv subjects

Zhihao Peng

Publications and source records attributed to Zhihao Peng.

17 recordsLinked to original sources
arXiv2026

NeuroClaw Technical Report

Agentic artificial intelligence systems promise to accelerate scientific workflows, but neuroimaging poses unique challenges: heterogeneous modalities (sMRI, fMRI, dMRI, EEG), long multi-stage pipelines, and persistent reproducibility risks. To address this gap, we present NeuroClaw, a domain-specialized multi-agent research assistant for executable and reproducible neuroimaging research. NeuroClaw operates directly on raw neuroimaging data across formats and modalities, grounding decisions in dataset semantics and BIDS metadata so users need not prepare curated inputs or bespoke model code. The platform combines harness engineering with end-to-end environment management, including pinned Python environments, Docker support, automated installers for common neuroimaging tools, and GPU configuration. In practice, this layer emphasizes checkpointing, post-execution verification, structured audit traces, and controlled runtime setup, making toolchains more transparent while improving reproducibility and auditability. A three-tier skill/agent hierarchy separates user-facing interaction, high-level orchestration, and low-level tool skills to decompose complex workflows into safe, reusable units. Alongside the NeuroClaw framework, we introduce NeuroBench, a system-level benchmark for executability, artifact validity, and reproducibility readiness. Across multiple multimodal LLMs, NeuroClaw-enabled runs yield consistent and substantial score improvements compared with direct agent invocation. Project homepage: https://cuhk-aim-group.github.io/NeuroClaw/index.html

cs.CV↗
arXiv2026

UniField: A Unified Field-Aware MRI Enhancement Framework

Magnetic Resonance Imaging (MRI) field-strength enhancement holds immense value for both clinical diagnostics and advanced research. However, existing methods typically focus on isolated enhancement tasks, such as specific 64mT-to-3T or 3T-to-7T transitions using limited subject cohorts, thereby failing to exploit the shared degradation patterns inherent across different field strengths and severely restricting model generalization. To address this challenge, we propose \methodname, a unified framework integrating multiple modalities and enhancement tasks to mutually promote representation learning by exploiting these shared degradation characteristics. Specifically, our main contributions are threefold. Firstly, to overcome MRI data scarcity and capture continuous anatomical structures, \methodname departs from conventional methods that treat 3D MRI volumes as independent 2D slices. Instead, we directly exploit comprehensive 3D volumetric information by leveraging pre-trained 3D foundation models, thereby embedding generalized and robust structural representations to significantly boost enhancement performance. In addition, to mitigate the spectral bias of mainstream flow-matching models that often over-smooth high-frequency details, we explicitly incorporate the physical mechanisms of magnetic fields to introduce a Field-Aware Spectral Rectification Mechanism (FASRM), tailoring customized spectral corrections to distinct field strengths. Finally, to resolve the fundamental data bottleneck, we organize and publicly release a comprehensive paired multi-field MRI dataset, which is an order of magnitude larger than existing datasets. Extensive experiments demonstrate our method's superiority over state-of-the-art approaches, achieving an average improvement of approximately 1.81 dB in PSNR and 9.47% in SSIM. Codes and datasets are available at: https://github.com/linyiyang98/UniField.

cs.CV↗
arXiv2026

Brain-WM: Brain Glioblastoma World Model

Precise prognostic modeling of glioblastoma (GBM) under varying treatment interventions is essential for optimizing clinical outcomes. While generative AI has shown promise in simulating GBM evolution, existing methods typically treat interventions as static conditional inputs rather than dynamic decision variables. Consequently, they fail to capture the complex, reciprocal interplay between tumor evolution and treatment response. To bridge this gap, we present Brain-WM, a pioneering brain GBM world model that unifies next-step treatment prediction and future MRI generation, thereby capturing the co-evolutionary dynamics between tumor and treatment. Specifically, Brain-WM encodes spatiotemporal dynamics into a shared latent space for joint autoregressive treatment prediction and flow-based future MRI generation. Then, instead of a conventional monolithic framework, Brain-WM adopts a novel Y-shaped Mixture-of-Transformers (MoT) architecture. This design structurally disentangles heterogeneous objectives, successfully leveraging cross-task synergies while preventing feature collapse. Finally, a synergistic multi-timepoint mask alignment objective explicitly anchors latent representations to anatomically grounded tumor structures and progression-aware semantics. Extensive validation on internal and external multi-institutional cohorts demonstrates the superiority of Brain-WM, achieving 91.5% accuracy in treatment planning and SSIMs of 0.8524, 0.8581, and 0.8404 for FLAIR, T1CE, and T2W sequences, respectively. Ultimately, Brain-WM offers a robust clinical sandbox for optimizing patient healthcare. The source code is made available at https://github.com/thibault-wch/Brain-GBM-world-model.

cs.CV↗
arXiv2026

Unveiling and Bridging the Functional Perception Gap in MLLMs: Atomic Visual Alignment and Hierarchical Evaluation via PET-Bench

While Multimodal Large Language Models (MLLMs) have demonstrated remarkable proficiency in tasks such as abnormality detection and report generation for anatomical modalities, their capability in functional imaging remains largely unexplored. In this work, we identify and quantify a fundamental functional perception gap: the inability of current vision encoders to decode functional tracer biodistribution independent of morphological priors. Identifying Positron Emission Tomography (PET) as the quintessential modality to investigate this disconnect, we introduce PET-Bench, the first large-scale functional imaging benchmark comprising 52,308 hierarchical QA pairs from 9,732 multi-site, multi-tracer PET studies. Extensive evaluation of 19 state-of-the-art MLLMs reveals a critical safety hazard termed the Chain-of-Thought (CoT) hallucination trap. We observe that standard CoT prompting, widely considered to enhance reasoning, paradoxically decouples linguistic generation from visual evidence in PET, producing clinically fluent but factually ungrounded diagnoses. To resolve this, we propose Atomic Visual Alignment (AVA), a simple fine-tuning strategy that enforces the mastery of low-level functional perception prior to high-level diagnostic reasoning. Our results demonstrate that AVA effectively bridges the perception gap, transforming CoT from a source of hallucination into a robust inference tool and improving diagnostic accuracy by up to 14.83%. Code and data are available at https://github.com/yezanting/PET-Bench.

cs.CV↗
arXiv2025

OmniBrainBench: A Comprehensive Multimodal Benchmark for Brain Imaging Analysis Across Multi-stage Clinical Tasks

Brain imaging analysis is crucial for diagnosing and treating brain disorders, and multimodal large language models (MLLMs) are increasingly supporting it. However, current brain imaging visual question-answering (VQA) benchmarks either cover a limited number of imaging modalities or are restricted to coarse-grained pathological descriptions, hindering a comprehensive assessment of MLLMs across the full clinical continuum. To address these, we introduce OmniBrainBench, the first comprehensive multimodal VQA benchmark specifically designed to assess the multimodal comprehension capabilities of MLLMs in brain imaging analysis with closed- and open-ended evaluations. OmniBrainBench comprises 15 distinct brain imaging modalities collected from 30 verified medical sources, yielding 9,527 validated VQA pairs and 31,706 images. It simulates clinical workflows and encompasses 15 multi-stage clinical tasks rigorously validated by a professional radiologist. Evaluations of 24 state-of-the-art models, including open-source general-purpose, medical, and proprietary MLLMs, highlight the substantial challenges posed by OmniBrainBench. Experiments reveal that proprietary MLLMs like GPT-5 (63.37%) outperform others yet lag far behind physicians (91.35%), while medical ones show wide variance in closed- and open-ended VQA. Open-source general-purpose MLLMs generally trail but excel in specific tasks, and all ones fall short in complex preoperative reasoning, revealing a critical visual-to-clinical gap. OmniBrainBench establishes a new standard to assess MLLMs in brain imaging analysis, highlighting the gaps against physicians. We publicly release our benchmark at link.

cs.CV↗
arXiv2025

Expertise-aware Multi-LLM Recruitment and Collaboration for Medical Decision-Making

Medical Decision-Making (MDM) is a complex process requiring substantial domain-specific expertise to effectively synthesize heterogeneous and complicated clinical information. While recent advancements in Large Language Models (LLMs) show promise in supporting MDM, single-LLM approaches are limited by their parametric knowledge constraints and static training corpora, failing to robustly integrate the clinical information. To address this challenge, we propose the Expertise-aware Multi-LLM Recruitment and Collaboration (EMRC) framework to enhance the accuracy and reliability of MDM systems. It operates in two stages: (i) expertise-aware agent recruitment and (ii) confidence- and adversarial-driven multi-agent collaboration. Specifically, in the first stage, we use a publicly available corpus to construct an LLM expertise table for capturing expertise-specific strengths of multiple LLMs across medical department categories and query difficulty levels. This table enables the subsequent dynamic selection of the optimal LLMs to act as medical expert agents for each medical query during the inference phase. In the second stage, we employ selected agents to generate responses with self-assessed confidence scores, which are then integrated through the confidence fusion and adversarial validation to improve diagnostic reliability. We evaluate our EMRC framework on three public MDM datasets, where the results demonstrate that our EMRC outperforms state-of-the-art single- and multi-LLM methods, achieving superior diagnostic performance. For instance, on the MMLU-Pro-Health dataset, our EMRC achieves 74.45% accuracy, representing a 2.69% improvement over the best-performing closed-source model GPT- 4-0613, which demonstrates the effectiveness of our expertise-aware agent recruitment strategy and the agent complementarity in leveraging each LLM's specialized capabilities.

cs.AI↗
arXiv2025

MAC: Masked Agent Collaboration Boosts Large Language Model Medical Decision-Making

Large language models (LLMs) have proven effective in artificial intelligence, where the multi-agent system (MAS) holds considerable promise for healthcare development by achieving the collaboration of LLMs. However, the absence of a systematic pipeline for agent construction and the rigidity of static collaboration patterns render current MAS-based models vulnerable to collaboration failures, resulting in substantial performance degradation in medical decision-making scenarios. To this end, we propose a novel Masked Agent Collaboration (MAC) framework that harnesses Pareto-optimal agent construction and cross-consistency maximization mechanisms to achieve adaptive progressive propagation of collaborative information, boosting the medical decision-making capacity. Specifically, we first conduct a Pareto-frontier factors analysis towards the LLMs pool to consider their key factors, including the model size, inference time, diversity score, and throughput ratio, where we calculate the similarity between pairwise outputs within an LLM to derive its diversity score. Beyond this analysis, we enable the identification of Pareto-optimal models that balance efficiency and capability, which are subsequently selected as collaborative agents to consider the fundamental trade-offs inherent in practical LLM deployment. Afterward, we measure the pairwise similarity between the outputs from collaborative agents to determine their cross-consistency values, subsequently masking out the agent with the lowest cross-consistency value to eliminate the output that is likely semantically inconsistent. Finally, we conduct collaboration of agents by achieving adaptive progressive propagation, where each agent aggregates the outputs of unmasked agents from the previous layer as its input to generate the corresponding output via prompt engineering.

cs.AI↗
arXiv2025

MCTS-Refined CoT: High-Quality Fine-Tuning Data for LLM-Based Repository Issue Resolution

LLMs demonstrate strong performance in auto-mated software engineering, particularly for code generation and issue resolution. While proprietary models like GPT-4o achieve high benchmarks scores on SWE-bench, their API dependence, cost, and privacy concerns limit adoption. Open-source alternatives offer transparency but underperform in complex tasks, especially sub-100B parameter models. Although quality Chain-of-Thought (CoT) data can enhance reasoning, current methods face two critical flaws: (1) weak rejection sampling reduces data quality, and (2) inadequate step validation causes error accumulation. These limitations lead to flawed reasoning chains that impair LLMs'ability to learn reliable issue resolution. The paper proposes MCTS-REFINE, an enhanced Monte Carlo Tree Search (MCTS)-based algorithm that dynamically validates and optimizes intermediate reasoning steps through a rigorous rejection sampling strategy, generating high-quality CoT data to improve LLM performance in issue resolution tasks. Key innovations include: (1) augmenting MCTS with a reflection mechanism that corrects errors via rejection sampling and refinement, (2) decomposing issue resolution into three subtasks-File Localization, Fault Localization, and Patch Generation-each with clear ground-truth criteria, and (3) enforcing a strict sampling protocol where intermediate outputs must exactly match verified developer patches, ensuring correctness across reasoning paths. Experiments on SWE-bench Lite and SWE-bench Verified demonstrate that LLMs fine-tuned with our CoT dataset achieve substantial improvements over baselines.Notably, Qwen2.5-72B- Instruct achieves 28.3%(Lite) and 35.0%(Verified) resolution rates, surpassing SOTA baseline SWE-Fixer-Qwen-72B with the same parameter scale, which only reached 24.7%(Lite) and 32.8%(Verified).

cs.SE↗
arXiv2025

Towards a general-purpose foundation model for fMRI analysis

Functional MRI (fMRI) is crucial for studying brain function and diagnosing neurological disorders. However, existing analysis methods suffer from reproducibility and transferability challenges due to complex preprocessing pipelines and task-specific model designs. In this work, we introduce NeuroSTORM (Neuroimaging Foundation Model with Spatial-Temporal Optimized Representation Modeling) that learns generalizable representations directly from 4D fMRI volumes and enables efficient transfer to diverse downstream applications. Specifically, NeuroSTORM is pre-trained on 28.65 million fMRI frames from over 50,000 subjects, spanning multiple centers and ages 5 to 100. It combines an efficient spatiotemporal modeling design and lightweight task adaptation to enable scalable pre-training and fast transfer to downstream applications. Here we show that NeuroSTORM consistently outperforms existing methods across five downstream tasks, including demographic prediction, phenotype prediction, disease diagnosis, re-identification, and state classification. On two multi-hospital clinical cohorts with 17 diagnoses, NeuroSTORM achieves the best diagnosis performance while remaining predictive of psychological and cognitive phenotypes. These results suggest that NeuroSTORM could become a standardized foundation model for reproducible and transferable fMRI analysis.

cs.CV↗
arXiv2025

EndoBench: A Comprehensive Evaluation of Multi-Modal Large Language Models for Endoscopy Analysis

Endoscopic procedures are essential for diagnosing and treating internal diseases, and multi-modal large language models (MLLMs) are increasingly applied to assist in endoscopy analysis. However, current benchmarks are limited, as they typically cover specific endoscopic scenarios and a small set of clinical tasks, failing to capture the real-world diversity of endoscopic scenarios and the full range of skills needed in clinical workflows. To address these issues, we introduce EndoBench, the first comprehensive benchmark specifically designed to assess MLLMs across the full spectrum of endoscopic practice with multi-dimensional capacities. EndoBench encompasses 4 distinct endoscopic scenarios, 12 specialized clinical tasks with 12 secondary subtasks, and 5 levels of visual prompting granularities, resulting in 6,832 rigorously validated VQA pairs from 21 diverse datasets. Our multi-dimensional evaluation framework mirrors the clinical workflow--spanning anatomical recognition, lesion analysis, spatial localization, and surgical operations--to holistically gauge the perceptual and diagnostic abilities of MLLMs in realistic scenarios. We benchmark 23 state-of-the-art models, including general-purpose, medical-specialized, and proprietary MLLMs, and establish human clinician performance as a reference standard. Our extensive experiments reveal: (1) proprietary MLLMs outperform open-source and medical-specialized models overall, but still trail human experts; (2) medical-domain supervised fine-tuning substantially boosts task-specific accuracy; and (3) model performance remains sensitive to prompt format and clinical task complexity. EndoBench establishes a new standard for evaluating and advancing MLLMs in endoscopy, highlighting both progress and persistent gaps between current models and expert clinical reasoning. We publicly release our benchmark and code.

cs.CV↗
arXiv2025

A Preliminary Investigation into Theory-Practice Barriers in Sino-New Zealand Undergraduate Computing Education

This paper investigates the barriers hindering the effective transition from theoretical knowledge to practical application in a Sino-New Zealand double-degree undergraduate computing program. In this unique educational setting, students study at a campus in China but complete both Chinese and New Zealand courses taught jointly by lecturers from both countries. Through a questionnaire administered to these students, we identify critical obstacles such as insufficient foundational knowledge, language barriers, cultural and pedagogical differences, and difficulties adapting to distinct educational systems. Our analysis reveals that these barriers significantly affect students' academic performance, engagement, and skill development. Based on the findings, we propose targeted interventions, including specialized bridging courses, enhanced language support, refined teaching methods, and improved resource allocation.

cs.CY↗
arXiv2025

MAP: Evaluation and Multi-Agent Enhancement of Large Language Models for Inpatient Pathways

Inpatient pathways demand complex clinical decision-making based on comprehensive patient information, posing critical challenges for clinicians. Despite advancements in large language models (LLMs) in medical applications, limited research focused on artificial intelligence (AI) inpatient pathways systems, due to the lack of large-scale inpatient datasets. Moreover, existing medical benchmarks typically concentrated on medical question-answering and examinations, ignoring the multifaceted nature of clinical decision-making in inpatient settings. To address these gaps, we first developed the Inpatient Pathway Decision Support (IPDS) benchmark from the MIMIC-IV database, encompassing 51,274 cases across nine triage departments and 17 major disease categories alongside 16 standardized treatment options. Then, we proposed the Multi-Agent Inpatient Pathways (MAP) framework to accomplish inpatient pathways with three clinical agents, including a triage agent managing the patient admission, a diagnosis agent serving as the primary decision maker at the department, and a treatment agent providing treatment plans. Additionally, our MAP framework includes a chief agent overseeing the inpatient pathways to guide and promote these three clinician agents. Extensive experiments showed our MAP improved the diagnosis accuracy by 25.10% compared to the state-of-the-art LLM HuatuoGPT2-13B. It is worth noting that our MAP demonstrated significant clinical compliance, outperforming three board-certified clinicians by 10%-12%, establishing a foundation for inpatient pathways systems.

cs.AI↗
arXiv2023

EGRC-Net: Embedding-induced Graph Refinement Clustering Network

Existing graph clustering networks heavily rely on a predefined yet fixed graph, which can lead to failures when the initial graph fails to accurately capture the data topology structure of the embedding space. In order to address this issue, we propose a novel clustering network called Embedding-Induced Graph Refinement Clustering Network (EGRC-Net), which effectively utilizes the learned embedding to adaptively refine the initial graph and enhance the clustering performance. To begin, we leverage both semantic and topological information by employing a vanilla auto-encoder and a graph convolution network, respectively, to learn a latent feature representation. Subsequently, we utilize the local geometric structure within the feature embedding space to construct an adjacency matrix for the graph. This adjacency matrix is dynamically fused with the initial one using our proposed fusion architecture. To train the network in an unsupervised manner, we minimize the Jeffreys divergence between multiple derived distributions. Additionally, we introduce an improved approximate personalized propagation of neural predictions to replace the standard graph convolution network, enabling EGRC-Net to scale effectively. Through extensive experiments conducted on nine widely-used benchmark datasets, we demonstrate that our proposed methods consistently outperform several state-of-the-art approaches. Notably, EGRC-Net achieves an improvement of more than 11.99\% in Adjusted Rand Index (ARI) over the best baseline on the DBLP dataset. Furthermore, our scalable approach exhibits a 10.73% gain in ARI while reducing memory usage by 33.73% and decreasing running time by 19.71%. The code for EGRC-Net will be made publicly available at \url{https://github.com/ZhihaoPENG-CityU/EGRC-Net}.

cs.LG↗
arXiv2022

Deep Attention-guided Graph Clustering with Dual Self-supervision

Existing deep embedding clustering works only consider the deepest layer to learn a feature embedding and thus fail to well utilize the available discriminative information from cluster assignments, resulting performance limitation. To this end, we propose a novel method, namely deep attention-guided graph clustering with dual self-supervision (DAGC). Specifically, DAGC first utilizes a heterogeneity-wise fusion module to adaptively integrate the features of an auto-encoder and a graph convolutional network in each layer and then uses a scale-wise fusion module to dynamically concatenate the multi-scale features in different layers. Such modules are capable of learning a discriminative feature embedding via an attention-based mechanism. In addition, we design a distribution-wise fusion module that leverages cluster assignments to acquire clustering results directly. To better explore the discriminative information from the cluster assignments, we develop a dual self-supervision solution consisting of a soft self-supervision strategy with a triplet Kullback-Leibler divergence loss and a hard self-supervision strategy with a pseudo supervision loss. Extensive experiments validate that our method consistently outperforms state-of-the-art methods on six benchmark datasets. Especially, our method improves the ARI by more than 18.14% over the best baseline.

cs.CV↗
arXiv2022

Adaptive Attribute and Structure Subspace Clustering Network

Deep self-expressiveness-based subspace clustering methods have demonstrated effectiveness. However, existing works only consider the attribute information to conduct the self-expressiveness, which may limit the clustering performance. In this paper, we propose a novel adaptive attribute and structure subspace clustering network (AASSC-Net) to simultaneously consider the attribute and structure information in an adaptive graph fusion manner. Specifically, we first exploit an auto-encoder to represent input data samples with latent features for the construction of an attribute matrix. We also construct a mixed signed and symmetric structure matrix to capture the local geometric structure underlying data samples. Then, we perform self-expressiveness on the constructed attribute and structure matrices to learn their affinity graphs separately. Finally, we design a novel attention-based fusion module to adaptively leverage these two affinity graphs to construct a more discriminative affinity graph. Extensive experimental results on commonly used benchmark datasets demonstrate that our AASSC-Net significantly outperforms state-of-the-art methods. In addition, we conduct comprehensive ablation studies to discuss the effectiveness of the designed modules. The code will be publicly available at https://github.com/ZhihaoPENG-CityU.

cs.CV↗
arXiv2021

Attention-driven Graph Clustering Network

The combination of the traditional convolutional network (i.e., an auto-encoder) and the graph convolutional network has attracted much attention in clustering, in which the auto-encoder extracts the node attribute feature and the graph convolutional network captures the topological graph feature. However, the existing works (i) lack a flexible combination mechanism to adaptively fuse those two kinds of features for learning the discriminative representation and (ii) overlook the multi-scale information embedded at different layers for subsequent cluster assignment, leading to inferior clustering results. To this end, we propose a novel deep clustering method named Attention-driven Graph Clustering Network (AGCN). Specifically, AGCN exploits a heterogeneity-wise fusion module to dynamically fuse the node attribute feature and the topological graph feature. Moreover, AGCN develops a scale-wise fusion module to adaptively aggregate the multi-scale features embedded at different layers. Based on a unified optimization framework, AGCN can jointly perform feature learning and cluster assignment in an unsupervised fashion. Compared with the existing deep clustering methods, our method is more flexible and effective since it comprehensively considers the numerous and discriminative information embedded in the network and directly produces the clustering results. Extensive quantitative and qualitative results on commonly used benchmark datasets validate that our AGCN consistently outperforms state-of-the-art methods.

cs.CV↗
arXiv2021

Maximum Entropy Subspace Clustering Network

Deep subspace clustering networks have attracted much attention in subspace clustering, in which an auto-encoder non-linearly maps the input data into a latent space, and a fully connected layer named self-expressiveness module is introduced to learn the affinity matrix via a typical regularization term (e.g., sparse or low-rank). However, the adopted regularization terms ignore the connectivity within each subspace, limiting their clustering performance. In addition, the adopted framework suffers from the coupling issue between the auto-encoder module and the self-expressiveness module, making the network training non-trivial. To tackle these two issues, we propose a novel deep subspace clustering method named Maximum Entropy Subspace Clustering Network (MESC-Net). Specifically, MESC-Net maximizes the entropy of the affinity matrix to promote the connectivity within each subspace, in which its elements corresponding to the same subspace are uniformly and densely distributed. Furthermore, we design a novel framework to explicitly decouple the auto-encoder module and the self-expressiveness module. We also theoretically prove that the learned affinity matrix satisfies the block-diagonal property under the independent subspaces. Extensive quantitative and qualitative results on commonly used benchmark datasets validate MESC-Net significantly outperforms state-of-the-art methods.

cs.CV↗