arXiv subjects

Yiuming Cheung

Publications and source records attributed to Yiuming Cheung.

5 recordsLinked to original sources

arXiv2026

Multi-Level Contextual Token Relation Modeling for Machine-Generated Text Detection

Machine-generated texts (MGTs) pose risks such as disinformation and phishing, underscoring the need for reliable detection. Metric-based methods, which extract statistically distinguishable features of MGTs, are often more practical than complex model-based methods that are prone to overfitting. Given their diverse designs, we first place representative metric-based methods within a unified framework, enabling a clear assessment of their advantages and limitations. Our analysis identifies a core challenge across these methods: the token-level detection score is easily biased by the inherent randomness of the MGTs generation process. Then, we theoretically derive the multi-hop transitions of the token-level detection score and explore their local and global relations. Based on these findings, we propose a multi-level contextual token relation modeling framework for MGT detection. Specifically, for local relations, we model them through a lightweight Markov-informed calibration module that refines token-level evidence before aggregation. For global relations, we introduce a rule-support reasoning module that uses explicit logical rules derived from contextual score statistics. Finally, we combine the local calibrated score and the global rule-support reasoning signal in a joint multi-level inference framework. Extensive experiments show broad and substantial improvements across various real-world scenarios, including cross-LLM and cross-domain settings, with low computational overhead.

cs.CL↗

arXiv2026

Adding Thermal Awareness to Visual Systems in Real-Time via Distilled Diffusion Models

Purely RGB-based vision models often fail to provide reliable cues in challenging scenarios such as nighttime and fog, leading to degraded performance and safety risks. Infrared imaging captures heat-emitting sources and provides critical complementary information, but existing high-fidelity fusion methods suffer from prohibitive latency, rendering them impractical for real-time edge deployment. To address this, we propose FusionProxy, a real-time image fusion module designed as a fully independent, plug-and-play component with diffusion level quality. FusionProxy exploits two complementary statistics of a teacher sample ensemble: per-pixel variance in raw image space, used to weight pixel-level supervision, and per-pixel variance inside frozen foundation backbones, used to route feature-level alignment spatially. Once trained, FusionProxy can be directly integrated into any visual perception system without joint optimization. Extensive experiments demonstrate that our method achieves superior performance on static recognition tasks and significantly enhances robustness in dynamic tasks, including closed-loop autonomous driving. Crucially, FusionProxy achieves real-time inference speeds on diverse platforms, from high-end GPUs to commodity hardware, providing a flexible and generalizable solution for all-day perception.

cs.CV↗

arXiv2026

Bringing Multimodal Large Language Models to Infrared-Visible Image Fusion Quality Assessment

Infrared-Visible image fusion (IVIF) aims to integrate thermal information and detailed spatial structures into a single fused image to enhance perception. However, existing evaluation approaches tend to over-optimize both hand-crafted no-reference statistics and full-reference metrics that treat the source images as pseudo ground truths. Recent IVIF reward-modelling efforts learn from human ratings but use scalar regression on aggregated scores, neither leveraging the reasoning of Multimodal Large Language Models (MLLMs) nor encoding per-image perceptual ambiguity in their supervision, but naively introducing MLLMs with discrete one-hot supervision likewise collapses fused images of similar quality into different rating levels. To address this, we introduce FuScore, which utilizes an MLLM to mimic human visual perception by producing continuous quality score, rather than discrete level predictions, enabling fine-grained discrimination among fused images of similar quality. We exploit the agreement among four IVIF-specific sub-dimensions to construct a per-image soft label whose sharpness reflects how consensual the overall judgment is. We further introduce a tripartite objective combining per-image distributional supervision, within-source-pair Thurstone fidelity for method-level ordering, and cross-source-pair Thurstone fidelity for scene-level ordering across scenes. Extensive experiments demonstrate that FuScore achieves state-of-the-art correlation with human visual preferences.

cs.CV↗

arXiv2017

Competitive and Penalized Clustering Auto-encoder

The paper has been withdrawn since more effective experiments should be completed. Auto-encoders (AE) has been widely applied in different fields of machine learning. However, as a deep model, there are a large amount of learnable parameters in the AE, which would cause over-fitting and slow learning speed in practice. Many researchers have been study the intrinsic structure of AE and showed different useful methods to regularize those parameters. In this paper, we present a novel regularization method based on a clustering algorithm which is able to classify the parameters into different groups. With this regularization, parameters in a given group have approximate equivalent values and over-fitting problem could be alleviated. Moreover, due to the competitive behavior of clustering algorithm, this model also overcomes some intrinsic problems of clustering algorithms like the determination of number of clusters. Experiments on handwritten digits recognition verify the effectiveness of our novel model.

cs.LG↗

arXiv2006

Topological fractal networks introduced by mixed degree distribution

Several fundamental properties of real complex networks, such as the small-world effect, the scale-free degree distribution, and recently discovered topological fractal structure, have presented the possibility of a unique growth mechanism and allow for uncovering universal origins of collective behaviors. However, highly clustered scale-free network, with power-law degree distribution, or small-world network models, with exponential degree distribution, are not self-similarity. We investigate networks growth mechanism of the branching-deactivated geographical attachment preference that learned from certain empirical evidence of social behaviors. It yields high clustering and spectrums of degree distribution ranging from algebraic to exponential, average shortest path length ranging from linear to logarithmic. We observe that the present networks fit well with small-world graphs and scale-free networks in both limit cases (exponential and algebraic degree distribution respectively), obviously lacking self-similar property under a length-scale transformation. Interestingly, we find perfect topological fractal structure emerges by a mixture of both algebraic and exponential degree distributions in a wide range of parameter values. The results present a reliable connection among small-world graphs, scale-free networks and topological fractal networks, and promise a natural way to investigate universal origins of collective behaviors.

physics.data-an↗