Graph-structured data underpins many critical applications. While foundation models have transformed language and vision via large-scale pretraining and lightweight adaptation, extending this paradigm to general, real-world graphs is challenging. In this work, we present Graph Billion-Foundation-Fusion (GraphBFF): an end-to-end recipe for building billion-parameter Graph Foundation Models (GFMs) for large-scale heterogeneous graphs. Central to the recipe is the GraphBFF Transformer, a flexible and scalable architecture designed for practical billion-scale GFMs. Using the GraphBFF, we present neural scaling laws for heterogeneous graphs and show that loss decreases predictably as either model capacity or training data scales, depending on which factor is the bottleneck. The GraphBFF framework provides concrete methodologies for data batching, pretraining, and fine-tuning for building GFMs at scale. We demonstrate the effectiveness of the framework over a real-world billion-scale graph, with an evaluation of a billion-parameter GraphBFF Transformer following the proposed recipe. Across ten diverse, real-world downstream tasks on graphs unseen during training, spanning node- and link-level classification and regression, GraphBFF consistently outperforms baselines, with large margins of up to 31 PRAUC points, including in few-shot settings. Finally, we discuss key challenges and open opportunities for making GFMs a practical and principled foundation for graph learning at industrial scale.
CausalMoE: A Billion-Scale Multimodal Foundation Model for Granger Causal Discovery with Pattern-Routed Heterogeneous Experts
State Key Laboratory of General Artificial Intelligence, School of Intelligence Science and Technology, Peking University（北京大学智能科学与技术学院通用人工智能国家重点实验室）；National Institute of Health Data Science, and Institute for Artificial Intelligence, Peking University（北京大学健康医疗大数据国家研究院、人工智能研究院）
UniTraj: Learning a Universal Trajectory Foundation Model from Billion-Scale Worldwide Traces
Southern University of Science and Technology（南方科技大学）；City University of Hong Kong（香港城市大学）；The Hong Kong University of Science and Technology (Guangzhou)（香港科技大学（广州））；Harbin Institute of Technology, Shenzhen（哈尔滨工业大学（深圳））
Macro Graph of Experts for Billion-Scale Multi-Task Recommendation
Jinan University（济南大学）；The Hong Kong Polytechnic University（香港理工大学）；City University of Macau（澳门城市大学）；Alibaba Group（阿里巴巴集团）；Beijing Institute of Technology（北京理工大学）；National University of Defense Technology（国防科技大学）；Beihang University（北航）
Construction and Applications of Billion-Scale Pre-Trained Multimodal Business Knowledge Graph
National University of Singapore（新加坡国立大学）；Alibaba Group（阿里巴巴集团）；Zhejiang University（浙江大学）；University of Manchester（曼彻斯特大学）；University of Edinburgh（爱丁堡大学）
Pixie: A System for Recommending 3+ Billion Items to 200+ Million Users in Real-Time
Pinterest
现代内容发现应用中的用户体验关键取决于高质量的个性化推荐。然而，由于大量的项目池、大量的用户以及对响应于用户动作并按需实时的【形】生成的推荐的要求，构建提供这种推荐的系统提出了重大挑战。在这里，我们介绍了Pixie，一个可扩展的基于图的实时的【形】推荐系统，我们开发和部署在Pinterest。给定一组用户特定的引脚作为查询，Pixie会从数十亿个可能的引脚中实时的【形】选择与查询最相关的引脚。为了生成推荐，我们开发了Pixie Random Walk算法，该算法利用了30亿个节点和170亿条边的Pinterest对象图。实验表明，与以前基于Hadoop的生产系统相比，Pixie提供的推荐可以提高50%的用户参与度。此外，我们还开发了一种图修剪策略，使推荐性能提高了58%。最后，我们讨论Pixie的系统方面，其中单个服务器每秒执行1，200个推荐请求，延迟为60毫秒。今天，Pixie支持的系统贡献了Pinterest上超过80%的用户参与度。
Graph Optimization Foundation Model: Tokenizing Graph via A Language-Model Paradigm
Faculty of Engineering, The University of Hong Kong（香港大学工程学院）；Costello College of Business, George Mason University（乔治·马歇尔大学商学院）；Faculty of Business and Economics, The University of Hong Kong（香港大学商学院）；College of Engineering, University of California, Berkeley（加州大学伯克利分校工程学院）
FIMP: Foundation Model-Informed Message Passing for Graph Neural Networks
Yale University（耶鲁大学）；Brown University（布朗大学）；New York University（纽约大学）；MIT（麻省理工学院）；ETH Zurich（苏黎世联邦理工学院）；Idaho State University（爱达荷州立大学）；University of Massachusetts Boston（马萨诸塞大学波士顿分校）；Baylor College of Medicine（贝勒医学院）；EPFL（洛桑联邦理工学院）；University of Southern California（南加州大学）
Position: Graph Foundation Models are Already Here
Michigan State University（密歇根州立大学）；Université de Montréal（蒙特利尔大学）；Mila - Québec AI Institute（米拉-魁北克人工智能研究所）；Rensselaer Polytechnic Institute（伦斯勒理工学院）；Snap Inc.（Snap公司）；Intel Labs（英特尔实验室）
VideoMAE V2: Scaling Video Masked Autoencoders with Dual Masking
State Key Laboratory for Novel Software Technology, Nanjing University（南京大学计算机软件新技术国家重点实验室）；Shanghai AI Lab（上海人工智能实验室）；Shenzhen Institute of Advanced Technology, CAS（中国科学院深圳先进技术研究院）
本文提出双重掩码策略和渐进式训练范式，高效扩展 VideoMAE 至十亿参数视频 ViT 模型，在多个视频基准上取得领先性能。
