Pai Chet Ng

Shanhong Liu, Pai Chet Ng, De Wen Soh et al.

Internet memes intertwine visual cues, textual content, and cultural context, making them particularly challenging to interpret in scenarios where humor, sarcasm, and harmful intent coexist. These complexities highlight the need for explainable meme understanding systems that can provide reliable and structured reasoning to support both accurate classification and human interpretability. However, existing multimodal classifiers either overlook these interdependencies or provide only limited interpretability. In this paper, we introduce MAR-12, a novel framework that leverages Vision Language Models (VLMs) for meme detection and understanding in settings where humorous and hateful elements may coexist. The framework first interprets each meme through twelve structured perspectives derived from humor and hate theories. It then applies a role-aware soft-gated attention mechanism to learn how much each perspective should contribute, followed by a prototype-based classifier for the final prediction. Finally, explanations are synthesized using both perspective-specific reasoning and learned attention weights, ensuring transparent and context-grounded justifications. We evaluate MAR-12 on the PrideMM and Memotion datasets, where it achieves up to 80.3% accuracy for humor detection and 75.9% accuracy for hate detection, outperforming state-of-the-art approaches. Furthermore, both human and GPT-4-based evaluations confirm that MAR-12 produces coherent and persuasive explanations, particularly for memes in which humorous and harmful cues co-occur.

Pai Chet Ng, Petros Spachos, James She et al.

This paper presents a nonlinear location estimation to infer the position of a user holding a smartphone. We consider a large location with $M$ number of grid points, each grid point is labeled with a unique fingerprint consisting of the received signal strength (RSS) values measured from $N$ number of Bluetooth Low Energy (BLE) beacons. Given the fingerprint observed by the smartphone, the user's current location can be estimated by finding the top-k similar fingerprints from the list of fingerprints registered in the database. Besides the environmental factors, the dynamicity in holding the smartphone is another source to the variation in fingerprint measurements, yet there are not many studies addressing the fingerprint variability due to dynamic smartphone positions held by human hands during online detection. To this end, we propose a nonlinear location estimation using the kernel method. Specifically, our proposed method comprises of two steps: 1) a beacon selection strategy to select a subset of beacons that is insensitive to the subtle change of holding positions, and 2) a kernel method to compute the similarity between this subset of observed signals and all the fingerprints registered in the database. The experimental results based on large-scale data collected in a complex building indicate a substantial performance gain of our proposed approach in comparison to state-of-the-art methods. The dataset consisting of the signal information collected from the beacons is available online.

Christos Korgialas, Gabriel Lee Jun Rong, Dion Jia Xu Ho et al.

The reliability of deepfake detectors frequently degrades under black-box adversarial transfer, as these models often rely on fragile, architecture-dependent forensic cues. Existing transfer attacks often lack semantic awareness and struggle to maintain effectiveness under strict no-query constraints, particularly when perturbations are transferred from convolutional surrogates to transformer-based targets. To address these limitations, this paper introduces ARMOR++, a robust multi-agent framework designed for high-transferability deepfake evasion. The framework leverages the Qwen2.5-VL Vision-Language Model (VLM) to supply spatial semantic priors, while the Qwen3 Large Language Model (LLM) orchestrates primitive selection, adaptive hyperparameter reparameterization, and entropy-regularized perturbation mixing. By integrating five complementary primitives, spanning dense optimization, saliency-based methods, spatial transformations, frequency-domain perturbations, and block-structured modifications, ARMOR++ effectively targets heterogeneous inductive biases. Rigorous evaluation on the AADD-2025 benchmark demonstrates that ARMOR++ significantly outperforms existing agentic and non-agentic baselines across both low- and high-quality image regimes. Statistical analysis confirms a substantial gain in blind-target Attack Success Rate (ASR) over the state-of-the-art agentic baseline, with further performance advantages evidenced against non-agentic benchmarks and under robust defensive configurations. These findings highlight a significant residual reliability gap in current deepfake detector deployments and demonstrate the efficacy of agentic orchestration in identifying latent vulnerabilities.

Shanhong Liu, Rui Cao, Pai Chet Ng et al.

Multimodal memes are dynamic and often require up to date background knowledge for interpretation. Existing methods often overlook such knowledge or rely on fixed parametric knowledge of pretrained models that may be incomplete, outdated, or unavailable for emerging memes. We introduce Query Retrieve Conclude, a zero shot framework that identifies missing knowledge, retrieves open web evidence, and synthesizes evidence grounded background knowledge for meme understanding and detection. We also introduce a curated meme understanding benchmark of recent memes from 2024 to 2026 with external background knowledge annotations. Experiments on three meme understanding datasets and five meme detection tasks show that our framework improves knowledge recovery, meme understanding and downstream detection over zero shot baselines.

Olivia Shanhong Liu, Pai Chet Ng, De Wen Soh et al.

Humorous memes blend visual and textual cues to convey irony, satire, or social commentary, posing unique challenges for AI systems that must interpret intent rather than surface correlations. Existing multimodal or prompting-based models generate explanations for humor but operate in an open loop,lacking the ability to critique or refine their reasoning once a prediction is made. We propose FLoReNce, an agentic feedback reasoning framework that treats meme understanding as a closed-loop process during learning and an open-loop process during inference. In the closed loop, a reasoning agent is critiqued by a judge; the error and semantic feedback are converted into control signals and stored in a feedback-informed, non-parametric knowledge base. At inference, the model retrieves similar judged experiences from this KB and uses them to modulate its prompt, enabling better, self-aligned reasoning without finetuning. On the PrideMM dataset, FLoReNce improves both predictive performance and explanation quality over static multimodal baselines, showing that feedback-regulated prompting is a viable path to adaptive meme humor understanding.

Gabriel Lee Jun Rong, Christos Korgialas, Dion Jia Xu Ho et al.

Existing automated attack suites operate as static ensembles with fixed sequences, lacking strategic adaptation and semantic awareness. This paper introduces the Agentic Reasoning for Methods Orchestration and Reparameterization (ARMOR) framework to address these limitations. ARMOR orchestrates three canonical adversarial primitives, Carlini-Wagner (CW), Jacobian-based Saliency Map Attack (JSMA), and Spatially Transformed Attacks (STA) via Vision Language Models (VLM)-guided agents that collaboratively generate and synthesize perturbations through a shared ``Mixing Desk". Large Language Models (LLMs) adaptively tune and reparameterize parallel attack agents in a real-time, closed-loop system that exploits image-specific semantic vulnerabilities. On standard benchmarks, ARMOR achieves improved cross-architecture transfer and reliably fools both settings, delivering a blended output for blind targets and selecting the best attack or blended attacks for white-box targets using a confidence-and-SSIM score.

Crystal Min Hui Poon, Pai Chet Ng, Xiaoxiao Miao et al.

Instruction-guided text-to-speech (TTS) research has reached a maturity level where excellent speech generation quality is possible on demand, yet two coupled biases persist: accent bias, where models default to dominant phonetic patterns, and linguistic bias, where dialect-specific lexical and cultural cues are ignored. These biases are interdependent, as authentic accent generation requires both accent fidelity and localized text. We present Contextual Linguistic Adaptation and Retrieval for Inclusive TTS sYnthesis (CLARITY), a backbone-agnostic framework that addresses these biases through dual-signal optimization: (i) contextual linguistic adaptation that localizes input text to the target dialect, and (ii) retrieval-augmented accent prompting (RAAP) that supplies accent-consistent speech prompts. Across twelve English accents, CLARITY improves accent accuracy and fairness while maintaining strong perceptual quality.

Jeremy Stephen Gabriel Yee, Pai Chet Ng, Zhengkui Wang et al.

This paper presents a systematic review of the infrastructure requirements for deploying Large Language Models (LLMs) on-device within the context of small and medium-sized enterprises (SMEs), focusing on both hardware and software perspectives. From the hardware viewpoint, we discuss the utilization of processing units like GPUs and TPUs, efficient memory and storage solutions, and strategies for effective deployment, addressing the challenges of limited computational resources typical in SME settings. From the software perspective, we explore framework compatibility, operating system optimization, and the use of specialized libraries tailored for resource-constrained environments. The review is structured to first identify the unique challenges faced by SMEs in deploying LLMs on-device, followed by an exploration of the opportunities that both hardware innovations and software adaptations offer to overcome these obstacles. Such a structured review provides practical insights, contributing significantly to the community by enhancing the technological resilience of SMEs in integrating LLMs.

Dion J. X. Ho, Gabriel Lee Jun Rong, Niharika Shrivastava et al.

We present MS-GAGA (Metric-Selective Guided Adversarial Generation Attack), a two-stage framework for crafting transferable and visually imperceptible adversarial examples against deepfake detectors in black-box settings. In Stage 1, a dual-stream attack module generates adversarial candidates: MNTD-PGD applies enhanced gradient calculations optimized for small perturbation budgets, while SG-PGD focuses perturbations on visually salient regions. This complementary design expands the adversarial search space and improves transferability across unseen models. In Stage 2, a metric-aware selection module evaluates candidates based on both their success against black-box models and their structural similarity (SSIM) to the original image. By jointly optimizing transferability and imperceptibility, MS-GAGA achieves up to 27% higher misclassification rates on unseen detectors compared to state-of-the-art attacks.

Javier Si Zhao Hong, Timothy Zoe Delaya, Sherwyn Chan Yin Kit et al.

This paper presents our approach to the first Multimodal Personality-Aware Depression Detection Challenge, focusing on multimodal depression detection using machine learning and deep learning models. We explore and compare the performance of XGBoost, transformer-based architectures, and large language models (LLMs) on audio, video, and text features. Our results highlight the strengths and limitations of each type of model in capturing depression-related signals across modalities, offering insights into effective multimodal representation strategies for mental health prediction.

Pai Chet Ng, Petros Spachos, Stefano Gregori et al.

Digital contact tracing has emerged as a viable tool supplementing manual contact tracing. To date, more than 100 contact tracing applications have been published to slow down the spread of highly contagious Covid-19. Despite subtle variabilities among these applications, all of them achieve contact tracing by manipulating the following three components: a) use a personal device to identify the user while designing a secure protocol to anonymize the user's identity; b) leverage networking technologies to analyze and store the data; c) exploit rich sensing features on the user device to detect the interaction among users and thus estimate the exposure risk. This paper reviews the current digital contact tracing based on these three components. We focus on two personal devices that are intimate to the user: smartphones and wearables. We discuss the centralized and decentralized networking approaches that use to facilitate the data flow. Lastly, we investigate the sensing feature available on smartphones and wearables to detect the proximity between any two users and present experiments comparing the proximity sensing performance between these two personal devices.

Pai Chet Ng, Petros Spachos, Stefano Gregori et al.

Businesses planning for the post-pandemic world are looking for innovative ways to protect the health and welfare of their employees and customers. Wireless technologies can play a key role in assisting contact tracing to quickly halt a local infection outbreak and prevent further spread. In this work, we present a wearable proximity and exposure notification solution based on a smartwatch that also promotes safe physical distancing in business, hospitality, or recreational facilities. Our proximity-based privacy-preserving contact tracing (P$^3$CT) leverages the Bluetooth Low Energy (BLE) technology for reliable proximity sensing, and an ambient signature protocol for preserving identity. Proximity sensing exploits the received signal strength (RSS) to detect the user's interaction and thus classifying them into low- or high-risk with respect to a patient diagnosed with an infectious disease. More precisely, a user is notified of their exposure based on their interactions, in terms of distance and time, with a patient. Our privacy-preserving protocol uses the ambient signatures to ensure that users' identities be anonymized. We demonstrate the feasibility of our proposed solution through extensive experimentation.

Pai Chet Ng, Petros Spachos, Konstantinos Plataniotis

Contact tracing is of paramount importance when it comes to preventing the spreading of infectious diseases. Contact tracing is usually performed manually by authorized personnel. Manual contact tracing is an inefficient, error-prone, time-consuming process of limited utility to the population at large as those in close contact with infected individuals are informed hours, if not days, later. This paper introduces an alternative way to manual contact tracing. The proposed Smart Contact Tracing (SCT) system utilizes the smartphone's Bluetooth Low Energy (BLE) signals and machine learning classifier to accurately and quickly determined the contact profile. SCT's contribution is two-fold: a) classification of the user's contact as high/low-risk using precise proximity sensing, and b) user anonymity using a privacy-preserving communications protocol. SCT leverages BLE's non-connectable advertising feature to broadcast a signature packet when the user is in the public space. Both broadcasted and observed signatures are stored in the user's smartphone and they are only uploaded to a secure signature database when a user is confirmed by public health authorities to be infected. Using received signal strength (RSS) each smartphone estimates its distance from other user's phones and issues real-time alerts when social distancing rules are violated. The paper includes extensive experimentation utilizing real-life smartphone positions and a comparative evaluation of five machine learning classifiers. Reported results indicate that a decision tree classifier outperforms other states of the art classification methods in terms of accuracy. Lastly, to facilitate research in this area, and to contribute to the timely development of advanced solutions the entire data set of six experiments with about 123,000 data points is made publicly available.