To address the unsustainable rise in public health expenditures, the Hong Kong SAR Government is shifting its strategic focus to primary healthcare and encouraging citizens to use community resources to self-manage their health. However, official clinical guidelines are fragmented across disparate departments and formats, creating significant access barriers. While general-purpose Large Language Models (LLMs) such as ChatGPT and DeepSeek offer potential solutions for information accessibility, they are prone to generating factually inaccurate content due to a lack of localized and domain-specific knowledge. To this end, we propose a Retrieval-Augmented Generation-Enhanced LLM system as Primary Healthcare Assistant (PriHA) in Hong Kong. Specifically, a tri-stage pipeline is proposed that leverages a query optimizer to generalize user intent-oriented sub-queries, followed by a novel Dual Retrieval Augmented Generation (DRAG) architecture for mixed-source retrieval and context-reorganized generation. Comprehensive experiments and a detailed case study demonstrate that our proposed method can outperform both ablations and baseline in terms of accuracy and clarity. Our research provides
Multilingual understanding is crucial for the cross-cultural applicability of Large Language Models (LLMs). However, evaluation benchmarks designed for Hong Kong's unique linguistic landscape, which combines Traditional Chinese script with Cantonese as the spoken form and its cultural context, remain underdeveloped. To address this gap, we introduce HKMMLU, a multi-task language understanding benchmark that evaluates Hong Kong's linguistic competence and socio-cultural knowledge. The HKMMLU includes 26,698 multi-choice questions across 66 subjects, organized into four categories: Science, Technology, Engineering, and Mathematics (STEM), Social Sciences, Humanities, and Other. To evaluate the multilingual understanding ability of LLMs, 90,550 Mandarin-Cantonese translation tasks were additionally included. We conduct comprehensive experiments on GPT-4o, Claude 3.7 Sonnet, and 18 open-source LLMs of varying sizes on HKMMLU. The results show that the best-performing model, DeepSeek-V3, struggles to achieve an accuracy of 75\%, significantly lower than that of MMLU and CMMLU. This performance gap highlights the need to improve LLMs' capabilities in Hong Kong-specific language and knowl
This paper presents the development of HKGAI-V1, a foundational sovereign large language model (LLM), developed as part of an initiative to establish value-aligned AI infrastructure specifically tailored for Hong Kong. Addressing the region's unique multilingual environment (Cantonese, Mandarin, and English), its distinct socio-legal context under the "one country, two systems" framework, and specific local cultural and value considerations, the model is built upon the DeepSeek architecture and systematically aligned with regional norms through a multifaceted full parameter fine-tuning process. It is further integrated with a retrieval-augmented generation (RAG) system to ensure timely and factually grounded information access. The core contribution lies in the design and implementation of a comprehensive, region-specific AI alignment and safety framework, demonstrated through two key achievements: 1) The successful development of HKGAI-V1 itself - which outper-forms general-purpose models in handling Hong Kong-specific culturally sensitive queries, and embodies a "governance-embedded" approach to digital sovereignty - empowers Hong Kong to exercise control over AI applications in
The Hongkong and Shanghai Banking Co (HSBC) just survived a civil war intermitted by World War II. By the 1950s, it obviously needed to close all its branches in Mao's People's Republic of China, yet could somehow hold its Shanghai branch, which continued likely in the shadows, as non-state banking was illegalised and even simple land owners were executed merely for being labelled "capitalist". This Asia-focused bank --in spite of it all-- grew from these conditions into the behemoth it is today. Part of the growth was based on the economic boom in its core market Hong Kong, to which HSBC likely also contributed. To expand and diversify, HSBC continued the growth strategy that already started since its early days in the 1860s, this time just also inorganically: It acquired other banks, in most cases fully and in other regions. The most important acquisition was the takeover of the roughly equally-sized UK-based Midland Bank; for the following reasons: 1) It came just a year after the 1991 change of HSBC's headquarters and place of incorporation to London, so HSBC could smoothly integrate with Midland. This step also came with an additional listing of securities in London, providing
Translating Hong Kong Court Judgments from English to Traditional Chinese is mandated by Articles 8-9 of the Basic Law, yet remains constrained by a shortage of parallel resources and rigorous demands on legal terminology, citation format, and judicial style. We introduce HKCFA Judgment 97-22, the first large-scale sentence-aligned parallel corpus for HK case law, comprising 344 professionally translated judgments (11,099 sentence pairs; 2.1M tokens) spanning 1997-2022. Building on this resource, we propose TransLaw, a multi-agent framework that decomposes translation into word-level expression, sentence-level translation, and multidimensional review, integrating a specialized Hong Kong legal glossary database, Retrieval-Augmented Generation, and iterative feedback, with four-dimensional expert review covering semantic alignment, terminology, citation, and style. Benchmarking 13 open-source and commercial LLMs, we demonstrate that TransLaw significantly outperforms single-agent baselines across all evaluated models, with convergence within 3 iterations. Human evaluation by 10 certified legal translators using our proposed Legal ACS metric confirms gains in legal-semantic accuracy,
Hong Kong' senior geography curriculum has included GIS since the early 2000s. However, GIS in secondary schools does not play a significant role in Hong Kong secondary geography education. Analyzing GIS benefits by literature review, it is believed that GIS should be included in both the senior and junior geography curriculum. Moreover, the literature review indicates that without clear instruction from the Hong Kong Education Bureau (EDB), low preparedness of Hong Kong geography teachers, and unsupportive attitudes from academia and textbook publishers, GIS cannot be implemented in secondary schools of Hong Kong. Therefore, suggestions are made for the EDB, geography teachers, academia and textbook publishers to facilitate GIS involvement in senior and junior geography curriculums. The EDB can develop clear guidelines for teachers, academia and textbook publishers' references, and offer student-centered GIS educational courses for teachers. It is important for teachers to be prepared for advanced GIS technology and to even learn along with students. Academics and textbook publishers can provide free GIS maps targeted at Hong Kong' junior and senior geography curriculums. Although
In the Hong Kong Observatory, the Analogue Forecast System (AFS) for precipitation has been providing useful reference in predicting possible daily rainfall scenarios for the next 9 days, by identifying historical cases with similar weather patterns to the latest output from the deterministic model of the European Centre for Medium-Range Weather Forecasts (ECMWF). Recent advances in machine learning allow more sophisticated models to be trained using historical data and the patterns of high-impact weather events to be represented more effectively. As such, an enhanced AFS has been developed using the deep learning technique autoencoder. The datasets of the fifth generation of the ECMWF Reanalysis (ERA5) are utilised where more meteorological elements in higher horizontal, vertical and temporal resolutions are available as compared to the previous ECMWF reanalysis products used in the existing AFS. The enhanced AFS features four major steps in generating the daily rain class forecasts: (1) preprocessing of gridded ERA5 and ECMWF model forecast, (2) feature extraction by the pretrained autoencoder, (3) application of optimised feature weightings based on historical cases, and (4) cal
Employee turnover is a critical challenge in financial markets, yet little is known about the role of professional networks in shaping career moves. Using the Hong Kong Securities and Futures Commission (SFC) public register (2007-2024), we construct temporal networks of 121,883 professionals and 4,979 firms to analyze and predict employee departures. We introduce a graph-based feature propagation framework that captures peer influence and organizational stability. Our analysis shows a contagion effect: professionals are 23% more likely to leave when over 30% of their peers depart within six months. Embedding these network signals into machine learning models improves turnover prediction by 30% over baselines. These results highlight the predictive power of temporal network effects in workforce dynamics, and demonstrate how network-based analytics can inform regulatory monitoring, talent management, and systemic risk assessment.
We present the first study of the Public Register of Licensed Persons and Registered Institutions maintained by the Hong Kong Securities and Futures Commission (SFC) through the lens of complex network analysis. This dataset, spanning 21 years with daily granularity, provides a unique view of the evolving social network between licensed professionals and their affiliated firms in Hong Kong's financial sector. Leveraging large language models, we classify firms (e.g., asset managers, banks) and infer the likely nationality and gender of employees based on their names. This application enhances the dataset by adding rich demographic and organizational context, enabling more precise network analysis. Our preliminary findings reveal key structural features, offering new insights into the dynamics of Hong Kong's financial landscape. We release the structured dataset to enable further research, establishing a foundation for future studies that may inform recruitment strategies, policy-making, and risk management in the financial industry.
As the global economic environment becomes increasingly unstable, enhancing financial flexibility to cope with risks has become the consensus of many companies. At the same time, environmental, social, and governance (ESG) performance may be one of the effective ways. We studied the impact of a firm's ESG performance on its financial flexibility with a sample of companies listed on the Hong Kong stock market from 2018 to 2022. The empirical results show that good environmental, social and governance performance can significantly improve a firm's financial flexibility. In addition, this paper also finds that the influence of ESG performance on financial flexibility is weak for state-owned enterprises due to the influence of governance structure and market characteristics. Finally, the further analysis shows that there is a mediating role played by financing constraints in this process. This study can provide background information for state-owned enterprises' governance, information disclosure, and corporate operations. It also has guiding significance for relevant investors, management and officials.
We study how legislation that restricts speech can induce online self-censorship and alter online discourse, using the recent Hong Kong national security law as a case study. We collect a dataset of 7 million historical Tweets from Hong Kong users, supplemented with historical snapshots of Tweet streams collected by other researchers. We compare online activity before and after enactment of the national security law, and we find that Hong Kong users demonstrate two types of self-censorship. First, Hong Kong users are more likely than a control group, sampled randomly from historical snapshots of Tweet streams, to remove past online activity. Specifically, Hong Kong users are over a third more likely than the control group to delete or restrict their account and over twice as likely to delete past posts. Second, we find that Hong Kong users post less often about politically sensitive topics that have been censored on social media in mainland China. This trend continues to increase.
Camera traps are used by ecologists globally as an efficient and non-invasive method to monitor animals. While it is time-consuming to manually label the collected images, recent advances in deep learning and computer vision has made it possible to automating this process [1]. A major obstacle to this is the generalisability of these models when applying these images to independently collected data from other parts of the world [2]. Here, we use a deep active learning workflow [3], and train a model that is applicable to camera trap images collected in Hong Kong.
This paper aims at comparing the time when Hong Kong universities used to ban ChatGPT to the current periods where it has become integrated in the academic processes. Bolted by concerns of integrity and ethical issues in technologies, institutions have adapted by moving towards the center adopting AI literacy and responsibility policies. This study examines new paradigms which have been developed to help implement these positives while preventing negative effects on academia. Keywords: ChatGPT, Academic Integrity, AI Literacy, Ethical AI Use, Generative AI in Education, University Policy, AI Integration in Academia, Higher Education and Technology
This study delves into the media analysis of China's ambitious Belt and Road Initiative (BRI), which, in a polarized world, and furthermore, owing to the very polarizing nature of the initiative itself, has received both strong criticisms and conversely positive coverage in media from across the world. In that context, Hong Kong's dynamic media environment, with a particular focus on its drastically changing press freedom before and after the implementation of the National Security Law is of further interest. Leveraging data science techniques, this study employs Global Database of Events, Language, and Tone (GDELT) to comprehensively collect and analyse (English) news articles on the BRI. Through sentiment analysis, we uncover patterns in media coverage over different periods from several countries across the globe, and delve further to investigate the the media situation in the Hong Kong region. This work thus provides valuable insights into how the Belt and Road Initiative has been portrayed in the media and its evolving reception on the global stage, with a specific emphasis on the unique media landscape of Hong Kong. In an era characterised by increasing globalisation and inte
Analysis and extraction of useful information from legal judgments using computational linguistics was one of the earliest problems posed in the domain of information retrieval. Presently, several commercial vendors exist who automate such tasks. However, a crucial bottleneck arises in the form of exorbitant pricing and lack of resources available in analysis of judgements mete out by Hong Kong's Legal System. This paper attempts to bridge this gap by providing several statistical, machine learning, deep learning and zero-shot learning based methods to effectively analyze legal judgments from Hong Kong's Court System. The methods proposed consists of: (1) Citation Network Graph Generation, (2) PageRank Algorithm, (3) Keyword Analysis and Summarization, (4) Sentiment Polarity, and (5) Paragrah Classification, in order to be able to extract key insights from individual as well a group of judgments together. This would make the overall analysis of judgments in Hong Kong less tedious and more automated in order to extract insights quickly using fast inferencing. We also provide an analysis of our results by benchmarking our results using Large Language Models making robust use of the H
The voting system in the Legislative Council of Hong Kong (Legco) is sometimes unicameral and sometimes bicameral, depending on whether the bill is proposed by the Hong Kong government. Therefore, although without any representative within Legco, the Hong Kong government has certain degree of legislative power --- as if there is a virtual representative of the Hong Kong government within the Legco. By introducing such a virtual representative of the Hong Kong government, we show that Legco is a three-dimensional voting system. We also calculate two power indices of the Hong Kong government through this virtual representative and consider the $C$-dimension and the $W$-dimension of Legco. Finally, some implications of this Legco model to the current constitutional reform in Hong Kong will be given.
We utilize a fundamentally different model of trading costs to look at the effect of the opening of the Hong Kong Shanghai Connect that links the stock exchanges in the two cities, arguably the biggest event in international business and finance since Christopher Columbus set sail for India. We design a novel methodology that compensates for the lack of data on trading costs in China. We estimate trading costs across similar positions on the dual listed set of securities in Hong Kong and China, hoping to provide useful pieces of information to help scale 'The Great Wall of Chinese Securities Trading Costs'. We then compare actual and estimated trading costs on a sample of real orders across the Hong Kong securities in the dual listed pair to establish the accuracy of our measurements. The primary question we seek to address is 'Which market would be better to trade to gain exposure to the same (or similar) set of securities or sectors?' We find that trading costs on Shanghai, which might have been lower than Hong Kong, might have become higher leading up to the Connect. What remains to be seen is whether this increase in trading costs is a temporary equilibrium due to the frenzy to
People are likely to engage in collective behaviour online during extreme events, such as the COVID-19 crisis, to express their awareness, actions and concerns. Hong Kong has implemented stringent public health and social measures (PHSMs) to curb COVID-19 epidemic waves since the first COVID-19 case was confirmed on 22 January 2020. People are likely to engage in collective behaviour online during extreme events, such as the COVID-19 crisis, to express their awareness, actions and concerns. Here, we offer a framework to evaluate interactions among individuals emotions, perception, and online behaviours in Hong Kong during the first two waves (February to June 2020) and found a strong correlation between online behaviours of Google search and the real-time reproduction numbers. To validate the model output of risk perception, we conducted 10 rounds of cross-sectional telephone surveys from February 1 through June 20 in 2020 to quantify risk perception levels over time. Compared with the survey results, the estimates of the risk perception of individuals using our network-based mechanistic model capture 80% of the trend of people risk perception (individuals who worried about being i
Thermal state reconstruction--reversing convection to recover the thermal structure of the mantle at an earlier geologic time--is an important tool to understand the evolution of mantle convection and its relation to seismic tomographic images and observations at the surface. Thermal state reconstructions are computationally expensive. Here we transformed the basic computational element, numerical solvers, into neural operators, a class of machine learning models for learning mappings between function spaces. Focusing on a specific architecture, Fourier Neural Operators, we demonstrate that they can represent not only a surrogate model like the Stokes system of equations using a purely physics informed approach, but also discover operators without explicit mathematical formulations or even ill-posedness from data, including the direct mapping between two convecting thermal states separated by a long time interval much larger than the Courant-Friedrichs-Lewy condition and its reversal. These neural operators significantly accelerate forward and inverse convection modelling by transforming forward physical processes into surrogate models with lower complexity while utilizing auto-dif
3D point cloud models suffer significant performance degradation under distribution shifts caused by sensor noise, occlusions, and environmental changes. Test-time adaptation (TTA) has emerged as a practical paradigm for mitigating this issue during inference. Recently, leveraging multi-view augmentation has shown promise in improving 3D TTA performance. However, existing multi-view approaches are often constrained by sequential optimization that treats each view independently. This sequential optimization leads to substantial inference latency due to repetitive optimization steps, making real-time adaptation impractical. To address this, we propose Masked Multi-View Test-Time Adaptation (MAMVI), which replaces sequential optimization with a unified single-step adaptation. Specifically, MAMVI utilizes a hybrid masking strategy that combines fixed ratios for stability with Beta-distributed sampling for diversity. By aggregating losses across multiple views, MAMVI performs adaptation through a single backward pass based on multi-view consensus. Additionally, a confidence-based adaptive learning rate is used to dynamically adjust the adaptation intensity for each sample. Extensive exp