Jun Pang

Multimedia-based recommendation is an essential tool of web applications, which has been successfully applied to e-commerce and content-sharing applications encompassing a vast amount of web multimedia content. This paper introduces a multimodal hierarchical graph contrastive learning architecture with diffusion-enhanced (MHDiff) for multimedia-based recommendation. Specifically, our MHDiff first designs a hierarchical modality-aware diffusion model to construct two multimodal hierarchical contrastive views: one at the semantic level and the other at the content level. For the semantic-level contrastive view, we incorporate semantic entities from items across different modalities to guide the generation of a user-item graph. For the content-level contrastive view, we integrate various modality features to guide the generation of a user-item graph. Consequently, our architecture combines a hierarchical modality-aware graph diffusion model with a hierarchical modality-aware contrastive learning paradigm. This approach effectively eliminates noise from multimodal content and enhance modality-aware user representation learning. Ultimately, a modality-aware main view is generated, leveraging the two multimodal hierarchical contrastive views to enhance node semantic information with minimal noise. Comprehensive experiments on multiple multimedia datasets demonstrate the superiority of our MHDiff on various state-of-the-art multimedia-based recommendation methods.

Recently, knowledge graphs have been utilised in recommendation systems to improve accuracy by integrating item-side auxiliary information. However, structural user-side knowledge is difficult to construct and integrate due to inherent scarcity and improper granularity. This paper introduces a graph contrastive learning with Semantic transitions-Enhanced DIffusion architecture based on Large Language Models (LLMs) for user-side knowledge-aware Recommendation (SEDIRec). Specifically, our SEDIRec first leverages LLMs to infer user interests from historical behaviors, integrating this user-side information with item-side and collaborative data to construct main views. Then, two contrastive views are generated using diffusion models with semantic transitions: one at the user-side level and the other at the item-side level. For both contrastive views, we integrate user-side or item-side information with collaborative data to generate a user-item graph. Subsequently, each user-item graph is transformed into collaborative data spaces via diffusion models for generating contrastive views. This procedure not only enhances the alignment between user/item-side information and the semantic spaces of collaborative data but also effectively eliminates noise. Extensive experiments on three datasets reveal the superiority of SEDIRec, especially for users with sparse interactions.

The rapid evolution of large language models (LLMs) has fuelled enthusiasm about their role in advancing scientific discovery, with studies exploring LLMs that autonomously generate and evaluate novel research ideas. However, little attention has been given to the possibility that such models could be exploited to produce harmful research by repurposing open science artefacts for malicious ends. We fill the gap by introducing an end-to-end pipeline that first bypasses LLM safeguards through persuasion-based jailbreaking, then reinterprets NLP papers to identify and repurpose their artefacts (datasets, methods, and tools) by exploiting their vulnerabilities, and finally assesses the safety of these proposals using our evaluation framework across three dimensions: harmfulness, feasibility of misuse, and soundness of technicality. Overall, our findings demonstrate that LLMs can generate harmful proposals by repurposing ethically designed open artefacts; however, we find that LLMs acting as evaluators strongly disagree with one another on evaluation outcomes: GPT-4.1 assigns higher scores (indicating greater potential harms, higher soundness and feasibility of misuse), Gemini-2.5-pro is markedly stricter, and Grok-3 falls between these extremes. This indicates that LLMs cannot yet serve as reliable judges in a malicious evaluation setup, making human evaluation essential for credible dual-use risk assessment.

Deep neural networks are widely deployed with quantization techniques to reduce memory and computational costs by lowering the numerical precision of their parameters. While quantization alters model parameters and their outputs, existing privacy analyses primarily focus on full-precision models, leaving a gap in understanding how bit-width reduction can affect privacy leakage. We present the first systematic study of the privacy-utility relationship in post-training quantization (PTQ), a versatile family of methods that can be applied to pretrained models without further training. Using membership inference attacks as our evaluation framework, we analyze three popular PTQ algorithms-AdaRound, BRECQ, and OBC-across multiple precision levels (4-bit, 2-bit, and 1.58-bit) on CIFAR-10, CIFAR-100, and TinyImageNet datasets. Our findings consistently show that low-precision PTQs can reduce privacy leakage. In particular, lower-precision models demonstrate up to an order of magnitude reduction in membership inference vulnerability compared to their full-precision counterparts, albeit at the cost of decreased utility. Additional ablation studies on the 1.58-bit quantization level show that quantizing only the last layer at higher precision enables fine-grained control over the privacy-utility trade-off. These results offer actionable insights for practitioners to balance efficiency, utility, and privacy protection in real-world deployments.

Boolean networks provide robust, explainable, and predictive models of cellular dynamics, especially for cellular differentiation and fate decision processes. Yet, the construction of such models is extremely challenging, as it requires integrating prior knowledge with experimental observation of the transcriptome, potentially relating thousands of genes. We present a general methodology for integrating transcriptome data and prior knowledge on the underlying gene regulatory network in order to generate automatically ensembles of Boolean networks able to reproduce the modeled qualitative behavior. Our methodology builds on the software BoNesis, which implements the automatic construction of Boolean networks from a specification of their expected structural and dynamical properties. We show how to transform transcriptome data into such a qualitative specification, and then how to exploit the generated ensembles of Boolean networks for identifying families of candidate models, and for predicting robust cellular reprogramming targets. We illustrate the scalability and versatility of our overall approach with two applications: the modeling of hematopoiesis from single-cell RNA-Seq data, and modeling the differentiation of bone marrow stromal cells into adipocytes and osteoblasts from bulk RNA-seq time series data. For this latter case, we took advantage of ensemble modeling to predict combinations of reprogramming factors for trans-differentiation that are robust to model uncertainties due to variations in experimental replicates and choice of binarization method. Moreover, we performed an in silico assessment of the fidelity and efficiency of the reprogramming and conducted preliminary experimental validation.