News & Stories

人工智能（AI）圖像編輯及生成模型獲廣泛應用於圖像創作，然而其對抽象概念如感覺和氛圍等理解精準度一直存在局限，且多依賴純文字指令，較難準確表達複雜圖像意思，亦無法捕捉風格、材質或光影等效果。由香港科技大學（科大）馮諾依曼研究院院長兼計算機科學及工程學系講座教授賈佳亞教授帶領的團隊成功開發名為「DreamOmni2」的AI圖像生成和編輯器，不僅擁有卓越的多模態指令編輯和實體物件生成能力，更在抽象概念的理解和生成方面有重大突破，讓AI不僅能「看圖」，更能「理解圖意」，多方面表現優於同類型開源和閉源模型，為AI創作開啟無限可能。直擊缺陷：解鎖抽象概念的創作潛能近年，圖像編輯及生成模型的發展進入爆發階段，新品頻出，但暫時仍未有任何模型能徹底克服實際操作上遇到的兩大缺陷。其一是文本指令的局限性，純文字指令難以準確描述人物特徵、抽象紋理等細節。其二是抽象概念的缺失，現有模型僅能處理有形實體，如人物、物件，無法有效應對髮型、妝容、紋理、光影效果或風格等抽象概念，極大程度上限制了創作空間。DreamOmni2則可解決有關難題，成功執行兩大全新任務，包括根據用家輸入的抽象或實體概念，執行多模態指令編輯和生成，真正成為「全能創作工具」。全面性能測試：超越現有開源與閉源模型在多模態指令生成任務中，DreamOmni2能基於圖片中的實體進行圖像生成，例如提取圖一的畫作掛在臥室牆上，將圖二盤子的材質套用在圖三的水杯，並將水杯放置在桌子上，以生成符合用家要求的新圖像（圖示一）。在多模態指令編輯任務中，DreamOmni2的表現亦非常優秀，例如將圖中帽子的顏色變成與另一張圖毛衣相同的配色（圖示二）。

2025年10月16日至19日，香港科大「創科行」（第4站）＠HICOOL 2025全球創業家高峰會系列活動成功舉辦！HICOOL全球創業家高峰會被譽為「全球創新創業風向標」，本屆高峰會吸引100餘個國家和地區的頂尖創新力量齊聚。作為「創科行」活動的核心連結場景，本屆高峰會活動為香港科大搭建了「成果展示+思想碰撞」的雙重平台，成為本次創科行的重頭戲。

The Hong Kong University of Science and Technology (HKUST) today launched SmartPath, a comprehensive artificial intelligence (AI) system designed to transform the entire pathology workflow for cancer care. The system was developed by an HKUST research team led by Prof. CHEN Hao, Director of Collaboration Center for Medical and Engineering Innovation, and Assistant Professor of the Department of Computer Science and Engineering and Department of Chemical and Biological Engineering. SmartPath provides integrated support for clinical diagnosis, subtyping, biomarker quantification, treatment response assessment, and prognostic follow-up across a wide spectrum of cancers, aiming to accelerate turnaround times and enhance the personalization of treatment plans.

A landmark study conducted by The Hong Kong University of Science and Technology (HKUST) has demonstrated that a genetic variant, TREM2 H157Y, significantly increases the risk of Alzheimer’s disease (AD) in individuals of ethnic Chinese descent. The research reveals that this variant confers a risk level comparable to that of the strongest known genetic risk factor for AD, APOE-ε4, and is associated with more rapid clinical progression and more severe neurodegeneration.This is the first in the field to conduct an in-depth family-based clinical case study on a Chinese-enriched genetic risk factor for AD, the TREM2 H157Y variant. The findings, published in Alzheimer’s & Dementia: The Journal of the Alzheimer’s Association, have profound implications for disease monitoring and patient management.

Membrane proteins are crucial for numerous biological processes and serve as important drug targets. For decades, scientists have relied on detergents to extract membrane proteins from cell membranes for structural studies. While detergent-based approaches have significantly advanced our understanding of membrane protein structures, they present certain limitations, such as resource-intensive detergent screening and the absence of native membrane lipids, which can hinder investigations into lipid-mediated regulation. To address these challenges, a research team led by Prof. DANG Shangyu from the Division of Life Science at the Hong Kong University of Science and Technology (HKUST) has developed a novel vesicle-based method that preserves the native lipid environment of membrane proteins, which can advance structural and functional studies.

A research team from the Hong Kong University of Science and Technology (HKUST), led by Professors Gan Jianping (Department of Ocean Science) and Professor Yang Can (Department of Mathematics), has developed a novel AI-powered tool named STIMP for diagnosing coastal ocean productivity and ecosystem health. STIMP introduces a novel paradigm that imputes missing data and then predicts Chlorophyll-a (Chl-a) concentrations across large spatiotemporal scales. In tests across four representative global coastal regions, STIMP significantly outperformed existing geoscience tools, reducing the mean absolute error (MAE) for imputation by up to 81.39% and for prediction by 58.99%. Accurate Chl-a prediction aids in early detection of harmful algal blooms, ecosystem protection, and provides data-driven insights for evidence-based policy-making.

A research team led by Prof. PAN Ding, Associate Professor from the Departments of Physics and Chemistry, and Dr. LI Shuo-Hui, Research Assistant Professor from the Department of Physics at the Hong Kong University of Science and Technology (HKUST), has developed a novel direct sampling method based on deep generative models. This method enables efficient sampling of the Boltzmann distribution across a continuous temperature range. The findings have been published in Physical Review Letters*.

The Hong Kong University of Science and Technology (HKUST) today received plaques from YIN Hejun, Minister of the Ministry of Science and Technology (MOST) for its three State Key Laboratories (SKLs) at the Government Headquarters of the Hong Kong Special Administrative Region (HKSAR). HKUST sincerely thanks MOST, and the Hong Kong Innovation, Technology, and Industry Bureau (ITIB) for their continued trust and recognition of the University’s research efforts. In addition to the two renamed SKLs following restructuring – the “State Key Laboratory of Nervous System Disorders (SKLNSD)” and the “State Key Laboratory of Displays and Opto-Electronics (SKLDOE)” – HKUST will also collaborate with The Hong Kong Polytechnic University (PolyU) to establish a new “State Key Laboratory of Climate Resilience for Coastal Cities (SKLCRCC)”.