Yawen Li | arXiv Science

arXiv2026

Graph Foundation Models for Recommendation: A Comprehensive Survey

Recommender systems (RS) serve as a fundamental tool for navigating the vast expanse of online information, with deep learning advancements playing an increasingly important role in improving ranking accuracy. Among these, graph neural networks (GNNs) excel at extracting higher-order structural information, while large language models (LLMs) are designed to process and comprehend natural language, making both approaches highly effective and widely adopted. Recent research has focused on graph foundation models (GFMs), which integrate the strengths of GNNs and LLMs to model complex RS problems more efficiently by leveraging the graph-based structure of user-item relationships alongside textual understanding. In this survey, we provide a comprehensive overview of GFM-based RS technologies by introducing a clear taxonomy of current approaches, diving into methodological details, and highlighting key challenges and future directions. By synthesizing recent advancements, we aim to offer valuable insights into the evolving landscape of GFM-based recommender systems.

cs.IR↗

arXiv2026

Towards Accurate and Lightweight Peripheral Neuroblastic Tumor Diagnosis via Contrastive Multi-scale Pathological Image Analysis

Peripheral neuroblastic tumors (pNTs) are among the most common extracranial solid tumors in children, and accurate pathological subtyping is important for risk stratification and treatment planning. However, pNT subtyping on hematoxylin-eosin whole-slide images (WSIs) remains challenging because of limited pediatric tumor cohorts, marked histological heterogeneity, inter-observer variability, and the computational burden of existing WSI classifiers. To address these challenges, we propose CoPath, a framework consisting of CoHisNet and PathVote. CoHisNet is a lightweight multi-scale feature-fusion network for patch-level histopathological classification. By replacing the multilayer perceptron components in Swin Transformer blocks and the classification head with Kolmogorov-Arnold Network layers, CoHisNet improves nonlinear feature modeling under a compact architecture. Its multi-scale interaction and contrast-driven feature-enhancement design enables the model to capture both tissue-level structures and fine-grained cellular morphology. PathVote further incorporates pathology-informed tissue-component priors to aggregate patch-level predictions into WSI-level decisions. We validated CoPath on a private two-branch PpNTs cohort and the public BreakHis breast cancer histopathology dataset. Experimental results show that CoPath achieves competitive or superior performance compared with general image classifiers, pathology foundation models under linear probing, and pathology-specific classification models, while maintaining substantially lower computational complexity. The source code is available at https://github.com/JSLiam94/CoPath.

cs.CV↗