Computer Sciences Commons^™

Freqllm: Frequency-Aware Large Language Models For Time Series Forecasting, Shunan Wang, Min Gao, Zongwei Wang, Yibing Bai, Feng Jiang, Guansong Pang

Research Collection School Of Computing and Information Systems

Large Language Models (LLMs) have recently shown promise in Time Series Forecasting (TSF) by effectively capturing intricate time-domain dependencies. However, our preliminary experiments reveal that standard LLM-based approaches often fail to capture global correlations, limiting predictive performance. We found that embedding frequency-domain signals smooths weight distributions and enhances structured correlations by clearly separating global trends (low-frequency components) from local variations (high-frequency components). Building on these insights, we propose FreqLLM, a novel framework that integrates frequency-domain semantic alignment into LLMs to refine prompts for improved time series analysis. By bridging the gap between frequency signals and textual embeddings, FreqLLM effectively captures …

Fine‑Tuning Multimodal Large Language Models For Product Bundling, Xiaohao Liu, Jie Wu, Zhulin Tao, Yunshan Ma, Yinwei Wei, Tat-Seng Chua

Research Collection School Of Computing and Information Systems

Recent advances in product bundling have leveraged multimodal information through sophisticated encoders, but remain constrained by limited semantic understanding and a narrow scope of knowledge. Therefore, some attempts employ In-context Learning (ICL) to explore the potential of large language models (LLMs) for their extensive knowledge and complex reasoning abilities. However, these efforts are inadequate in understanding mulitmodal data and exploiting LLMs' knowledge for product bundling. To bridge the gap, we introduce Bundle-MLLM, a novel framework that fine-tunes LLMs through a hybrid item tokenization approach within a well-designed optimization strategy. Specifically, we integrate textual, media, and relational data into a unified …

Anomalygfm: Graph Foundation Model For Zero/Few-Shot Anomaly Detection, Hezhe Qiao, Chaoxi Niu, Ling Chen, Guansong Pang

Research Collection School Of Computing and Information Systems

Graph anomaly detection (GAD) aims to identify abnormal nodes that differ from the majority of the nodes in a graph, which has been attracting significant attention in recent years. Existing generalist graph models have achieved remarkable success in different graph tasks but struggle to generalize to the GAD task. This limitation arises from their difficulty in learning generalized knowledge for capturing the inherently infrequent, irregular and heterogeneous abnormality patterns in graphs from different domains. To address this challenge, we propose AnomalyGFM, a GAD-oriented graph foundation model that supports zero-shot inference and few-shot prompt tuning for GAD in diverse graph datasets. …

Affinitytune: A Prompt-Tuning Framework For Few-Shot Anomaly Detection On Graphs, Jingyan Chen, Guanghui Zhu, Guansong Pang, Chunfeng Yuan, Yihua Huang

Research Collection School Of Computing and Information Systems

Graph anomaly detection (GAD) is a critical task with applications in domains such as networking, finance, and bioinformatics. % However, the scarcity of labeled anomalies and the limitations of unsupervised methods hinder effective detection. % While semi-supervised and few-shot learning approaches offer improvements, they struggle with knowledge transfer and rely heavily on labeled data. % Recent advancements in prompt tuning on graphs provide a promising direction, but their application to heterophilous graphs in anomaly detection remains underexplored. % In this work, we propose AffinityTune, a novel framework for few-shot graph anomaly detection based on prompt tuning. % Our approach introduces …

Llm2rec: Large Language Models Are Powerful Embedding Models For Sequential Recommendation, Yingzhi He, Xiaohao Liu, An Zhang, Yunshan Ma, Tat‑Seng Chua

Research Collection School Of Computing and Information Systems

Sequential recommendation aims to predict users' future interactions by modeling collaborative filtering (CF) signals from historical behaviors of similar users or items. Traditional sequential recommenders predominantly rely on ID-based embeddings, which capture CF signals through high-order co-occurrence patterns. However, these embeddings depend solely on past interactions, lacking transferable knowledge to generalize to unseen domains. Recent advances in large language models (LLMs) have motivated text-based recommendation approaches that derive item representations from textual descriptions. While these methods enhance generalization, they fail to encode CF signals-i.e., latent item correlations and preference patterns-crucial for effective recommendation. We argue that an ideal embedding model …

Gnncontext: Gnn-Based Code Context Prediction For Programming Tasks, Xiaoye Zheng, Zhiyuan Wan, Shun Liu, Kaiwen Yang, David Lo, Xiaohu Yang

Research Collection School Of Computing and Information Systems

A code context model comprises source code elements and their relations relevant to a programming task. The capture and use of code context models in software tools can benefit software development practices, such as code navigation and search. Prior research has explored approaches that leverage either the structural information of code or interaction histories of developers with integrated development environments to automate the construction of code context models. However, these approaches primarily capture shallow syntactic and lexical features of code elements, with limited ability to capture contextual and structural dependencies among neighboring code elements. In this paper, we propose GNNContext, …

Equivalence And Similarity Refutation For Probabilistic Programs, Krishnendu Chatterjee, Ehsan Kafshdar Goharshady, Petr Novotný, Dorde Zikelic

Research Collection School Of Computing and Information Systems

We consider the problems of statically refuting equivalence and similarity of output distributions defined by a pair of probabilistic programs. Equivalence and similarity are two fundamental relational properties of probabilistic programs that are essential for their correctness both in implementation and in compilation. In this work, we present a new method for static equivalence and similarity refutation. Our method refutes equivalence and similarity by computing a function over program outputs whose expected value with respect to the output distributions of two programs is different. The function is computed simultaneously with an upper expectation supermartingale and a lower expectation submartingale for …

Reimagining Education With Ai, Margherita Pagani, Steven M. Miller, Jerry Wind

Research Collection School Of Computing and Information Systems

This chapter examines AI’s transformative potential in education, focusing on Generative AI (GenAI) and Large Language Models (LLMs) while at the same time emphasizing the importance of grounding and guiding AI efforts with learning science and education research findings. It synthesizes analyses and expert recommendations, highlighting opportunities like personalized learning and enhanced teacher productivity, alongside challenges such as over-reliance on AI. Practical steps for instructors include adopting a question-first approach, utilizing AI for personalized feedback, designing AI-enhanced learning experiences, fostering critical thinking, and ensuring ethical AI use. The chapter concludes with strategic recommendations for leveraging AI to sustainably improve educational …

Dreamanime: Learning Style-Identity Textual Disentanglement For Anime And Beyond, Chenshu Xu, Yangyang Xu, Huaidong Zhang, Xuemiao Xu, Shengfeng He

Research Collection School Of Computing and Information Systems

Text-to-image generation models have significantly broadened the horizons of creative expression through the power of natural language. However, navigating these models to generate unique concepts, alter their appearance, or reimagine them in unfamiliar roles presents an intricate challenge. For instance, how can we exploit language-guided models to transpose an anime character into a different art style, or envision a beloved character in a radically different setting or role? This paper unveils a novel approach named DreamAnime, designed to provide this level of creative freedom. Using a minimal set of 2-3 images of a user-specified concept such as an anime character …

Machine Learning For Digital Biomarker-Based Detection Of Cognitive Decline, Seng Khoon The

Dissertations and Theses Collection (Open Access)

Dementia is a neurodegenerative disease with a prevalence rate expected to triple by 2050, posing a significant challenge for health services. To impede the increasing prevalence, medical professionals and scientists are actively investigating technology to detect cognitive decline at a reversible stage known as Mild Cognitive Impairment (MCI). Digital biomarker technology is an emerging pragmatic approach to permit objective, ecologically valid, and long-term continuous measurement of cognitive health status, rendering it as one of the promising technologies for early MCI detection. Despite its potential, it is nontrivial to encode, extract and combine predictive information from these digital biomarker technologies; advanced …

Enhancing Graph Representation Learning Through Self-Supervision: An Augmentation Perspective, Jianyuan Bo

Dissertations and Theses Collection (Open Access)

Graph representation learning has become fundamental in various domains, from social networks to molecular structures, enabling extraction of meaningful patterns from graph-structured data. While deep learning approaches, particularly graph neural networks, have shown promising results, their effectiveness is often limited by the scarcity of labeled data. This challenge is particularly acute in graph domains where annotation requires specialized expertise and is prohibitively expensive. Self-supervised learning has emerged as a promising direction to address this limitation by creating auxiliary tasks from unlabeled data, with augmentation strategies playing a crucial role in their success.

Current graph self-supervised learning methods face several critical …

Unambiguous Granularity Distillation For Asymmetric Image Retrieval, Hongrui Zhang, Yi Xie, Haoquan Zhang, Cheng Xu, Xuandi Luo, Donglei Chen, Xuemiao Xu, Huaidong Zhang, Pheng Ann Heng, Shengfeng He

Research Collection School Of Computing and Information Systems

Previous asymmetric image retrieval methods based on knowledge distillation have primarily focused on aligning the global features of two networks to transfer global semantic information from the gallery network to the query network. However, these methods often fail to effectively transfer local semantic information, limiting the fine-grained alignment of feature representation spaces between the two networks. To overcome this limitation, we propose a novel approach called Layered-Granularity Localized Distillation (GranDist). GranDist constructs layered feature representations that balance the richness of contextual information with the granularity of local features. As we progress through the layers, the contextual information becomes more detailed, …

Robust Relevance Feedback For Interactive Known-Item Video Search, Zhixin Ma, Chong-Wah Ngo

Research Collection School Of Computing and Information Systems

Known-item search (KIS) involves only a single search target, making relevance feedback-typically a powerful technique for efficiently identifying multiple positive examples to infer user intent-inapplicable. PicHunter addresses this issue by asking users to select the top-k most similar examples to the unique search target from a displayed set. Under ideal conditions, when the user's perception aligns closely with the machine's perception of similarity, consistent and precise judgments can elevate the target to the top position within a few iterations. However, in practical scenarios, expecting users to provide consistent judgments is often unrealistic, especially when the underlying embedding features used for …

Fashiondpo: Fine‑Tune Fashion Outfit Generation Model Using Direct Preference Optimization, Mingzhe Yu, Yunshan Ma, Lei Wu, Changshuo Wang, Xue Li, Lei Meng

Research Collection School Of Computing and Information Systems

Personalized outfit generation aims to construct a set of compatible and personalized fashion items as an outfit. Recently, generative AI models have received widespread attention, as they can generate fashion items for users to complete an incomplete outfit or create a complete outfit. However, they have limitations in terms of lacking diversity and relying on the supervised learning paradigm. Recognizing this gap, we propose a novel framework FashionDPO, which fine-tunes the fashion outfit generation model using direct preference optimization. This framework aims to provide a general fine-tuning approach to fashion generative models, refining a pre-trained fashion outfit generation model using …

Artificial Insights Or Historical Fidelity? Crafting An Ethical Framework For The Use Of Genai In The Restoration, Reconstruction And Recreation Of Movable Cultural Heritage, David Ocón, Chunzhi Yin, Jose Luna

Research Collection School of Social Sciences

This article explores the ethical considerations surrounding using Generative Artificial Intelligence (GenAI) in preserving movable cultural heritage, focusing specifically on its application in restoration, reconstruction, and recreation. While GenAI offers innovative methods for preserving and recreating cultural heritage, it also presents significant ethical challenges. The article reviews current studies on the role of GenAI in heritage preservation alongside relevant ethical guidelines and proposes a tailored ethical framework for its application in movable heritage. The framework addresses several critical ethical concerns, including cultural integrity and sensitivity, accuracy and authenticity, intellectual property rights, sustainability and social impact, and governance and ethical accountability. …

Evaluating Chatgpt To Answer Multi-Modal Exercises In Computer Science Education, Eng Lieh Ouh, Kar Way Tan, Siaw Ling Lo, Benjamin Gan

Research Collection School Of Computing and Information Systems

This study investigates ChatGPT-4o's ability to answer multi-modal assessment exercises in computer science (CS) courses. While the use of large language models (LLMs) to answer text-based exercises are extensively researched, their ability to answer exercises involving artifacts of other modalities remains underexplored. To close this gap, we evaluate ChatGPT-4o's answers to 120 multi-modal CS exercises in programming, software design, human-computer interaction, statistical analysis, process analysis, and simulation. The multi-modal artifacts in these exercises include class diagrams, sequence diagrams, user interface images, analytical charts, workflow diagrams and object-flow diagrams. Our comparisons to the expected answers of these exercises show that ChatGPT-4o …

Unbounded Multi-Hop Proxy Re-Encryption With Hra Security: An Lwe-Based Optimization, Xiaohan Wan, Yang Wang, Haiyang Xue, Mingqiang Wang

Research Collection School Of Computing and Information Systems

Proxy re-encryption (PRE) schemes enable a semi-honest proxy to transform a ciphertext of one user i to another user j while preserving the privacy of the underlying message. Multi-hop PRE schemes allow a legal ciphertext to undergo multiple transformations, but for lattice-based multi-hop PREs, the number of transformations is typically bounded due to the increase of error terms. Recently, Zhao et al. (ESORICS 2024) introduced a lattice-based unbounded multi-hop (homomorphic) PRE scheme that supports an unbounded number of hops. Nevertheless, their scheme only achieves the selective CPA security. In contrast, Fuchsbauer et al. (PKC 2019) proposed a generic framework for …

An Incentive Mechanism For Privacy Preserved Data Trading With Verifiable Data Disturbance, Man Zhang, Xinghua Li, Bin Luo, Yanbing Ren, Yinbin Miao, Ximeng Liu, Robert H. Deng

Research Collection School Of Computing and Information Systems

To motivate data owners’ (DOs’) trading willingness, the existing incentive mechanisms allow DOs to independently disturb data following data consumer's (DC’s) availability requirement. However, they cannot motivate DOs’ honest disturbance, which is attributed to DOs’ independent disturbance without any supervision. Thus, we implement an incentive mechanism for privacy preserved data trading with verifiable data disturbance where an honest-but-curious disturbance generator (DG) is additionally introduced to supervise DOs’ local disturbance and assist disturbance verification between DOs and DC. Specifically, DG generates the disturbance strategies and secretly distributes to DOs following private information retrieval, guaranteeing DOs's local disturbance's privacy and verifiability with …

Foodlmm: A Versatile Food Assistant Using Large Multi-Modal Model, Yuehao Yin, Huiyan Qi, Bin Zhu, Jingjing Chen, Yu-Gang Jiang, Chong-Wah Ngo

Research Collection School Of Computing and Information Systems

Large Multi-modal Models (LMMs) have made impressive progress in many vision-language tasks. Nevertheless, the performance of general LMMs in specific domains is still far from satisfactory. This paper proposes FoodLMM, a versatile food assistant based on LMMs with various capabilities, including food recognition, ingredient recognition, recipe generation, nutrition estimation, food segmentation and multi-round conversation. To facilitate FoodLMM to deal with tasks beyond pure text output, we introduce a series of novel task-specific tokens and heads, enabling the model to predict food nutritional values and multiple segmentation masks. We adopt a two-stage training strategy. In the first stage, we utilize multiple …

Query Understanding In Llm-Based Conversational Information Seeking, Yifei Yuan, Zahra Abbasiantaeb, Mohammad Aliannejadi, Yang Deng

Research Collection School Of Computing and Information Systems

Query understanding in CIS involves accurately interpreting user intent through context-aware interactions. This includes resolving ambiguities, refining queries, and adapting to evolving information needs. LLM enhance this process by interpreting nuanced language and adapting dynamically, improving the relevance and precision of search results in real-time. In this tutorial, we explore advanced techniques to enhance query understanding in LLM-based CIS systems. We delve into LLM-driven methods for developing robust evaluation metrics to assess query understanding quality in multi-turn interactions, strategies for building more interactive systems, and applications like proactive query management and query reformulation. We also discuss key challenges in integrating …

Modeling Multiple Tasks In Recommendation Systems, Dinh Hieu Do

Dissertations and Theses Collection (Open Access)

Traditional research in recommendation systems has largely centered on the static offline supervised learning setting. In this paradigm, all available user-item interaction data is collected and partitioned into fixed training, validation, and test sets. Models are developed and evaluated in this controlled environment, where the underlying data distribution is assumed to remain unchanged. This approach offers clear advantages: it simplifies experimentation, enables reproducible benchmarking, and allows for straightforward comparisons between algorithms.

However, this static offline setting does not reflect the realities faced by modern recommendation systems. In real-world applications, data is dynamic and ever-evolving, where new users and items are …

From Sparse Feedback To Sequential Decision-Making: Learning Safety Constraints With Weak Supervision, Siow Meng Low

Dissertations and Theses Collection (Open Access)

Real-world decision-making often involves safety constraints that are implicit, non-Markovian, or difficult to specify directly. Standard reinforcement learning (RL) approaches typically assume access to fully specified cost functions and constraint budgets—assumptions that limit their applicability in domains where such structure must instead be inferred from data. This dissertation develops a sequence of methods for learning safety-relevant structure from weak supervision, such as sparse binary feedback on trajectory segments, and using these signals to guide planning and policy optimization.

The first part of the dissertation introduces a sample-efficient method for planning in continuous Markov Decision Processes (MDPs) using deep reactive policies. …

Efficient Prompt Tuning For Hierarchical Ingredient Recognition, Yinxuan Gui, Bin Zhu, Jingjing Chen, Chong-Wah Ngo

Research Collection School Of Computing and Information Systems

Fine-grained ingredient recognition presents a significant challenge due to the diverse appearances of ingredients, resulting from different cutting and cooking methods. While existing approaches have shown promising results, they still require extensive training costs and focus solely on fine-grained ingredient recognition. In this paper, we address these limitations by introducing an efficient prompt-tuning framework that adapts pretrained visual-language models (VLMs), such as CLIP, to the ingredient recognition task without requiring full model finetuning. Additionally, we introduce three-level ingredient hierarchies to enhance both training performance and evaluation robustness. Specifically, we propose a hierarchical ingredient recognition task, designed to evaluate model performance …

O-Mapl: Offline Multi-Agent Preference Learning, The Viet Bui, Tien Mai, Hong Thanh Nguyen

Research Collection School Of Computing and Information Systems

Inferring reward functions from demonstrations is a key challenge in reinforcement learning (RL), particularly in multi-agent RL (MARL). The large joint state-action spaces and intricate inter-agent interactions in MARL make inferring the joint reward function especially challenging. While prior studies in single-agent settings have explored ways to recover reward functions and expert policies from human preference feedback, such studies in MARL remain limited. Existing methods typically combine two separate stages, supervised reward learning, and standard MARL algorithms, leading to unstable training processes. In this work, we exploit the inherent connection between reward functions and Q functions in cooperative MARL to …

Unified Neural Backdoor Removal With Only Few Clean Samples Through Unlearning And Relearning, Nay Myat Min, Long H. Pham, Jun Sun

Research Collection School Of Computing and Information Systems

Deep neural networks have achieved remarkable success across various applications; however, their vulnerability to backdoor attacks poses severe security risks—especially in situations where only a limited set of clean samples is available for defense. In this work, we address this critical challenge by proposing ULRL (UnLearn and ReLearn for backdoor removal), a novel two-phase approach for comprehensive backdoor removal. Our method first employs an unlearning phase, in which the network’s loss is intentionally maximized on a small clean dataset to expose neurons that are excessively sensitive to backdoor triggers. Subsequently, in the relearning phase, these suspicious neurons are recalibrated using …

Llmscan: Causal Scan For Llm Misbehavior Detection, Mengdi Zhang, Kai Kiat Goh, Peixin Zhang, Jun Sun, Lin Xin Rose, Hongyu Zhang

Research Collection School Of Computing and Information Systems

Despite the success of Large Language Models (LLMs) across various fields, their potential to generate untruthful and harmful responses poses significant risks, particularly in critical applications. This highlights the urgent need for systematic methods to detect and prevent such misbehavior. While existing approaches target specific issues such as harmful responses, this work introduces LLMSCAN, an innovative LLM monitoring technique based on causality analysis, offering a comprehensive solution. LLMSCAN systematically monitors the inner workings of an LLM through the lens of causal inference, operating on the premise that the LLM’s ‘brain’ behaves differently when generating harmful or untruthful responses. By analyzing …

Advancing Food Nutrition Estimation Via Visual-Ingredient Feature Fusion, Huiyan Qi, Bin Zhu, Chong-Wah Ngo, Jingjing Chen, Ee-Peng Lim

Research Collection School Of Computing and Information Systems

Nutrition estimation is an important component of promoting healthy eating and mitigating diet-related health risks. Despite advances in tasks such as food classification and ingredient recognition, progress in nutrition estimation is limited due to the lack of datasets with nutritional annotations. To address this issue, we introduce FastFood, a dataset with 84,446 images across 908 fast food categories, featuring ingredient and nutritional annotations. In addition, we propose a new model-agnostic Visual-Ingredient Feature Fusion (VIF2 ) method to enhance nutrition estimation by integrating visual and ingredient features. Ingredient robustness is improved through synonym replacement and resampling strategies during training. The ingredient-aware …

Unveiling Knowledge Boundary Of Large Language Models For Trustworthy Information Access, Yang Deng, Moxin Li, Liang Pang, Wenxuan Zhang, Wai Lam

Research Collection School Of Computing and Information Systems

Large Language Models (LLMs) have emerged as powerful tools for generating content and facilitating information seeking across diverse domains. While their integration into conversational systems opens new avenues for interactive information-seeking experiences, their effectiveness is constrained by their knowledge boundaries—the limits of what they know and their ability to provide reliable, truthful, and contextually appropriate information. Understanding these boundaries is essential for maximizing the utility of LLMs for real-time information seeking while ensuring their reliability and trustworthiness. In this tutorial, we will explore the taxonomy of knowledge boundary in LLMs, addressing their handling of uncertainty, response calibration, and mitigation of …

Repairing Adversarial Texts Through Perturbation, Guoliang Dong, Jingyi Wang, Jun Sun, Sudipta Chattopadhyay, Xinyu Wang, Ting Dai, Jie Shi, Jin Song Dong

Research Collection School Of Computing and Information Systems

It is known that neural networks are subject to attacks through adversarial perturbations. Worse yet, such attacks are impossible to eliminate, i.e., the adversarial perturbation is still possible after applying mitigation methods such as adversarial training. Multiple approaches have been developed to detect and reject such adversarial inputs. Rejecting suspicious inputs however may not be always feasible or ideal. First, normal inputs may be rejected due to false alarms generated by the detection algorithm. Second, denial-of-service attacks may be conducted by feeding such systems with adversarial inputs. To address this, in this work, we focus on the text domain and …

Leakage-Resilient Easily Deployable And Efficiently Searchable Encryption (Edese), Jiaming Yuan, Yingjiu Li, Jun Li, Daoyuan Wu, Jianting Ning, Yangguang Tian, Robert H. Deng

Research Collection School Of Computing and Information Systems

Easily Deployable and Efficiently Searchable Encryption (EDESE) is a cryptographic primitive designed for practical searchable applications, offering efficient search and easy deployment. However, it remains vulnerable to Leakage-Abuse attacks, allowing adversaries to exploit keyword-matching processes to extract sensitive information. To address these vulnerabilities, we introduce Leakage-Resilient EDESE (LR-EDESE) with k-indistinguishability and controlled leakage functions. We then propose Volume Leakage-Resilient EDESE (VLR-EDESE), a new scheme to protect against both query and document volume leakage. Our experimental results demonstrate that at k = 5000 (maximum security setting), VLR-EDESE incurs an overhead of 63× compared to the baseline EDESE without leakage protection, outperforming …