arXiv subjects

Wai Kin Victor Chan

Publications and source records attributed to Wai Kin Victor Chan.

7 recordsLinked to original sources

arXiv2026

Multi-level context Modeling for consistent expert selection in Mixture-of-Experts

Mixture-of-Experts (MoE) enables efficient scaling of Transformer models by routing tokens to a small subset of experts. However, existing routers typically condition expert selection on shallow or isolated token representations, which often produce unstable and semantically inconsistent routing decisions across layers. In this work, we revisit expert selection from a representation perspective and identify context incompleteness as a key bottleneck limiting effective expert specialization. To address this issue, we propose Multi-level Context Fusion MOE (MCF-MOE), a framework that constructs context-aware representations by integrating complementary signals from cross-layer semantic aggregation and local token-level interactions, enabling more informative and consistent expert selection. Experiments on language modeling and understanding benchmarks demonstrate that MCF-MOE consistently improves routing consistency and downstream performance over strong MoE baselines, highlighting the importance of contextual completeness in expert routing. The code is available at https://github.com/shuhanhuang/MCF-MOE.

cs.CL↗

arXiv2026

When the Prompt Becomes Visual: Vision-Centric Jailbreak Attacks for Large Image Editing Models

Recent advances in large image editing models have shifted the paradigm from text-driven instructions to vision-prompt editing, where user intent is inferred directly from visual inputs such as marks, arrows, and visual-text prompts. While this paradigm greatly expands usability, it also introduces a critical and underexplored safety risk: the attack surface itself becomes visual. In this work, we propose Vision-Centric Jailbreak Attack (VJA), the first visual-to-visual jailbreak attack that conveys malicious instructions purely through visual inputs. To systematically study this emerging threat, we introduce IESBench, a safety-oriented benchmark for image editing models. Extensive experiments on IESBench demonstrate that VJA effectively compromises state-of-the-art commercial models, achieving attack success rates of up to 80.9% on Nano Banana Pro and 70.1% on GPT-Image-1.5. To mitigate this vulnerability, we propose a training-free defense based on introspective multimodal reasoning, which substantially improves the safety of poorly aligned models to a level comparable with commercial systems, without auxiliary guard models and with negligible computational overhead. Our findings expose new vulnerabilities, provide both a benchmark and practical defense to advance safe and trustworthy modern image editing systems. Warning: This paper contains offensive images created by large image editing models.

cs.CV↗

arXiv2026

VPA-Guard: Defending and Benchmarking Image-to-Video Generation Against Visual Prompt Attacks

Recent advancements in Image-to-Video (I2V) generation have transformed input images from simple appearance references into interactive control interfaces where visual cues such as arrows, sketches, and emojis orchestrate complex video dynamics with unprecedented controllability. However, these seemingly innocuous static cues can be interpreted by models as executable temporal instructions, unfolding into harmful actions in the generated videos. Despite the severity of this threat, existing safety benchmarks remain predominantly focused on text-based and content-only image-based jailbreaks, leaving implicit visual prompt attacks insufficiently explored. To bridge this gap, we present VVA-Bench, the first systematic benchmark for evaluating video generation safety under categorized vision-centric prompt attacks. Extensive experiments on VVA-Bench demonstrate that state-of-the-art models are highly susceptible to such attacks, with Attack Success Rates (ASR) reaching 100.0\% on Wan 2.7 and 74.8\% on Veo 3.1. To mitigate these risks, we propose VPA-Guard, a retrieval-augmented and self-evolving defense framework. By leveraging few-shot reasoning to identify latent malicious intents, our method reduces the attack ASR by 44.2\% and the harmfulness score by 73.4\% on average, while maintaining the model's utility for legitimate user edits. Our work provides both a rigorous benchmark and an effective defense strategy to advance safe and socially responsible multimodal generation.

cs.CV↗

arXiv2023

CDR-Adapter: Learning Adapters to Dig Out More Transferring Ability for Cross-Domain Recommendation Models

Data sparsity and cold-start problems are persistent challenges in recommendation systems. Cross-domain recommendation (CDR) is a promising solution that utilizes knowledge from the source domain to improve the recommendation performance in the target domain. Previous CDR approaches have mainly followed the Embedding and Mapping (EMCDR) framework, which involves learning a mapping function to facilitate knowledge transfer. However, these approaches necessitate re-engineering and re-training the network structure to incorporate transferrable knowledge, which can be computationally expensive and may result in catastrophic forgetting of the original knowledge. In this paper, we present a scalable and efficient paradigm to address data sparsity and cold-start issues in CDR, named CDR-Adapter, by decoupling the original recommendation model from the mapping function, without requiring re-engineering the network structure. Specifically, CDR-Adapter is a novel plug-and-play module that employs adapter modules to align feature representations, allowing for flexible knowledge transfer across different domains and efficient fine-tuning with minimal training costs. We conducted extensive experiments on the benchmark dataset, which demonstrated the effectiveness of our approach over several state-of-the-art CDR approaches.

cs.IR↗

arXiv2022

A Constructive Heuristic Algorithm for 3D Bin Packing of Irregular Shaped Items

The three-dimensional bin packing problem (3D-BPP) plays an important role in city logistics and manufacturing environments, due to its direct relevance to operational cost. Most existing literature have investigated the conventional 3D-BPP, in which the shape of items are typically considered as regular shapes, e.g., rectangular-shaped rigid boxes or cylindrical-shaped containers. However, 3D-BPP for non-rectangular shaped items are quite common in varies delivery schemes, especially in fresh food delivery, and few published studies focusing on these issues. In this paper, we address a novel 3D-BPP variant in which the shape changing factor of non-rectangular and deformable items is incorporated to further enhance the loading efficiency and reduce the operational cost of related companies. Motivated by the compression process of item-loading, we propose a constructive heuristic (i.e., an improved dynamic-volume-based packing algorithm) to solve the studied problem. Experimental results over a set of randomly generated instances reveal that considering shape changing factor is indeed able to achieve higher space utilization than that of conventional schemes, thereby has potential to save packaging and delivering cost, as well as enhance operation efficiency.

cs.CG↗

arXiv2022

Reinforcement learning for multi-item retrieval in the puzzle-based storage system

Nowadays, fast delivery services have created the need for high-density warehouses. The puzzle-based storage system is a practical way to enhance the storage density, however, facing difficulties in the retrieval process. In this work, a deep reinforcement learning algorithm, specifically the Double&Dueling Deep Q Network, is developed to solve the multi-item retrieval problem in the system with general settings, where multiple desired items, escorts, and I/O points are placed randomly. Additionally, we propose a general compact integer programming model to evaluate the solution quality. Extensive numerical experiments demonstrate that the reinforcement learning approach can yield high-quality solutions and outperforms three related state-of-the-art heuristic algorithms. Furthermore, a conversion algorithm and a decomposition framework are proposed to handle simultaneous movement and large-scale instances respectively, thus improving the applicability of the PBS system.

cs.LG↗

arXiv2021

Learning Large-scale Location Embedding From Human Mobility Trajectories with Graphs

An increasing amount of location-based service (LBS) data is being accumulated and helps to study urban dynamics and human mobility. GPS coordinates and other location indicators are normally low dimensional and only representing spatial proximity, thus difficult to be effectively utilized by machine learning models in Geo-aware applications. Existing location embedding methods are mostly tailored for specific problems that are taken place within areas of interest. When it comes to the scale of a city or even a country, existing approaches always suffer from extensive computational cost and significant data sparsity. Different from existing studies, we propose to learn representations through a GCN-aided skip-gram model named GCN-L2V by considering both spatial connection and human mobility. With a flow graph and a spatial graph, it embeds context information into vector representations. GCN-L2V is able to capture relationships among locations and provide a better notion of similarity in a spatial environment. Across quantitative experiments and case studies, we empirically demonstrate that representations learned by GCN-L2V are effective. As far as we know, this is the first study that provides a fine-grained location embedding at the city level using only LBS records. GCN-L2V is a general-purpose embedding model with high flexibility and can be applied in down-streaming Geo-aware applications.

cs.SI↗