While large language models (LLMs) have become the de facto framework for literature-related tasks, they still struggle to function as domain-specific literature agents due to their inability to connect pieces of knowledge and reason across domain-specific contexts, terminologies, and nomenclatures. This challenge underscores the need for a tool that facilitates such domain-specific adaptation and enables rigorous benchmarking across literature tasks. To that end, we introduce LitBench, a benchmarking tool designed to enable the development and evaluation of domain-specific LLMs tailored to literature-related tasks. At its core, LitBench uses a data curation process that generates domain-specific literature sub-graphs and constructs training and evaluation datasets based on the textual attributes of the resulting nodes and edges. The tool is designed for flexibility, supporting the curation of literature graphs across any domain chosen by the user, whether high-level fields or specialized interdisciplinary areas. In addition to dataset curation, LitBench defines a comprehensive suite of literature tasks, ranging from node and edge level analyses to advanced applications such as related work generation. These tasks enable LLMs to internalize domain-specific knowledge and relationships embedded in the curated graph during training, while also supporting rigorous evaluation of model performance. Our results show that small domain-specific LLMs trained and evaluated on LitBench datasets achieve competitive performance compared to state-of-the-art models like GPT-4o and DeepSeek-R1. To enhance accessibility and ease of use, we open-source the tool along with an AI agent tool that streamlines data curation, model training, and evaluation.

Recent advances in spatial transcriptomics (ST) and cost reductions have enabled large-scale multi-slice ST data generation, enhancing the statistical power to detect subtle biological signals. However, cross-slice inconsistencies and data quality variability present significant analytical challenges. To overcome these limitations, we developed MUSE, a computational framework designed for multislice joint embedding, spatial domain identification, and gene expression imputation. Specifically, MUSE integrates a two-module architecture to ensure robust cross-slice alignment and data harmonization. The alignment module models each slice as a graph and employs optimal transport to align cells across slices while preserving spatial continuity. The optimization module further refines integration by incorporating an alignment loss, allowing lower-quality data to leverage structural information from higher-quality slices. Additionally, MUSE generates virtual neighbors from aligned cells, enriching contextual information and mitigating data sparsity. These design principles enable seamless integration with existing single-slice methods, extending their applicability to multi-slice ST analysis. To comprehensively evaluate its performance, we applied MUSE to 12 real and 48 simulated datasets spanning a range of data qualities. Across all metrics, MUSE consistently outperformed existing methods in cross-slice consistency, spatial domain identification, and gene expression imputation. To promote accessibility and adoption, we provide MUSE as an open-source software package. As multi-slice ST datasets become increasingly prevalent, MUSE provides a robust and extensible framework designed to effectively integrate growing numbers of slices, thereby advancing the analysis of tissue architectures and spatial gene expression in complex biological systems.

Hyperbolic neural networks have emerged as a powerful tool for modeling hierarchical data structures prevalent in real-world datasets. Notably, residual connections, which facilitate the direct flow of information across layers, have been instrumental in the success of deep neural networks. However, current methods for constructing hyperbolic residual networks suffer from limitations such as increased model complexity, numerical instability, and errors due to multiple mappings to and from the tangent space. To address these limitations, we introduce LResNet, a novel Lorentzian residual neural network based on the weighted Lorentzian centroid in the Lorentz model of hyperbolic geometry. Our method enables the efficient integration of residual connections in Lorentz hyperbolic neural networks while preserving their hierarchical representation capabilities. We demonstrate that our method can theoretically derive previous methods while offering improved stability, efficiency, and effectiveness. Extensive experiments on both graph and vision tasks showcase the superior performance and robustness of our method compared to state-of-the-art Euclidean and hyperbolic alternatives. Our findings highlight the potential of LResNet for building more expressive neural networks in hyperbolic embedding space as a generally applicable method to multiple architectures, including CNNs, GNNs, and graph Transformers.

Foundation models pre-trained on massive datasets, including large language models (LLMs), vision-language models (VLMs), and large multimodal models, have demonstrated remarkable success in diverse downstream tasks. However, recent studies have shown fundamental limitations of these models: (1) limited representational capacity(2) lower adaptability, and (3) diminishing scalability. These shortcomings raise a critical question: is Euclidean geometry truly the optimal inductive bias for all foundation models, or could incorporating alternative geometric spaces enable models to better align with the intrinsic structure of real-world data and improve reasoning processes? Hyperbolic spaces, a class of non-Euclidean manifolds characterized by exponential volume growth with respect to distance, offer a mathematically grounded solution. These spaces enable low-distortion embeddings of hierarchical structures (e.g., trees, taxonomies) and power-law distributions with substantially fewer dimensions compared to Euclidean counterparts. Recent advances have leveraged these properties to enhance foundation models, including improving LLMs' complex reasoning ability, VLMs' zero-shot generalization, and cross-modal semantic alignment, while maintaining parameter efficiency. This paper provides a comprehensive review of hyperbolic neural networks and their recent development for foundation models. We further outline key challenges and research directions to advance the field.

Generative models have shown great promise in collaborative filtering by capturing the underlying distribution of user interests and preferences. However, existing approaches struggle with inaccurate posterior approximations and misalignment with the discrete nature of recommendation data, limiting their expressiveness and real-world performance. To address these limitations, we propose FlowCF, a novel flow-based recommendation system leveraging flow matching for collaborative filtering. We tailor flow matching to the unique challenges in recommendation through two key innovations: (1) a behavior-guided prior that aligns with user behavior patterns to handle the sparse and heterogeneous user-item interactions, and (2) a discrete flow framework to preserve the binary nature of implicit feedback while maintaining the benefits of flow matching, such as stable training and efficient inference. Extensive experiments demonstrate that FlowCF achieves state-of-the-art recommendation accuracy across various datasets with the fastest inference speed, making it a compelling approach for real-world recommender systems. The code is available at https://github.com/chengkai-liu/FlowCF.

With the advent of large language models (LLMs), managing scientific literature via LLMs has become a promising direction of research. However, existing approaches often overlook the rich structural and semantic relevance among scientific literature, limiting their ability to discern the relationships between pieces of scientific knowledge, and suffer from various types of hallucinations. These methods also focus narrowly on individual downstream tasks, limiting their applicability across use cases. We propose LitFM, the first literature foundation model designed for a wide variety of practical downstream tasks on domain-specific literature, with a focus on citation information. At its core, LitFM contains a novel graph retriever that can provide accurate and diverse recommendations for LLM to integrate graph structure information and relevant literature. LitFM also leverages a knowledge-infused LLM, fine-tuned through a well-developed instruction paradigm. It enables LitFM to extract domain-specific knowledge from literature and reason relationships among them. By integrating citation graphs during both training and inference, LitFM can generalize to unseen papers and accurately assess their relevance within existing literature. Additionally, we introduce new large-scale literature citation benchmark datasets on three academic fields, featuring sentence-level citation information and local context. Extensive experiments validate the superiority of LitFM, achieving 28.1% improvement on retrieval task in precision, and an average improvement of 7.52% over state-of-the-art across six downstream literature-related tasks.

Hyperbolic geometry have shown significant potential in modeling complex structured data, particularly those with underlying tree-like and hierarchical structures. Despite the impressive performance of various hyperbolic neural networks across numerous domains, research on adapting the Transformer to hyperbolic space remains limited. Previous attempts have mainly focused on modifying self-attention modules in the Transformer. However, these efforts have fallen short of developing a complete hyperbolic Transformer. This stems primarily from: (i) the absence of well-defined modules in hyperbolic space, including linear transformation layers, LayerNorm layers, activation functions, dropout operations, etc. (ii) the quadratic time complexity of the existing hyperbolic self-attention module w.r.t the number of input tokens, which hinders its scalability. To address these challenges, we propose, Hypformer, a novel hyperbolic Transformer based on the Lorentz model of hyperbolic geometry. In Hypformer, we introduce two foundational blocks that define the essential modules of the Transformer in hyperbolic space. Furthermore, we develop a linear self-attention mechanism in hyperbolic space, enabling hyperbolic Transformer to process billion-scale graph data and long-sequence inputs for the first time. Our experimental results confirm the effectiveness and efficiency of \method across various datasets, demonstrating its potential as an effective and scalable solution for large-scale data representation and large models.

Subgraph counting is the problem of counting the occurrences of a given query graph in a large target graph. Large-scale subgraph counting is useful in various domains, such as motif analysis for social network and loop counting for money laundering detection. Recently, to address the exponential runtime complexity of scalable subgraph counting, neural methods are proposed. However, existing approaches fall short in three aspects. Firstly, the subgraph counts vary from zero to millions for different graphs, posing a much larger challenge than regular graph regression tasks. Secondly, current scalable graph neural networks have limited expressive power and fail to efficiently distinguish graphs for count prediction. Furthermore, existing neural approaches cannot predict query occurrence positions.

Temporal knowledge graphs (TKGs) are valuable resources for capturing evolving relationships among entities, yet they are often plagued by noise, necessitating robust anomaly detection mechanisms. Existing dynamic graph anomaly detection approaches struggle to capture the rich semantics introduced by node and edge categories within TKGs, while TKG embedding methods lack interpretability, undermining the credibility of anomaly detection. Moreover, these methods falter in adapting to pattern changes and semantic drifts resulting from knowledge updates. To tackle these challenges, we introduce AnoT, an efficient TKG summarization method tailored for interpretable online anomaly detection in TKGs. AnoT begins by summarizing a TKG into a novel rule graph, enabling flexible inference of complex patterns in TKGs. When new knowledge emerges, AnoT maps it onto a node in the rule graph and traverses the rule graph recursively to derive the anomaly score of the knowledge. The traversal yields reachable nodes that furnish interpretable evidence for the validity or the anomalous of the new knowledge. Overall, AnoT embodies a detector-updater-monitor architecture, encompassing a detector for offline TKG summarization and online scoring, an updater for real-time rule graph updates based on emerging knowledge, and a monitor for estimating the approximation error of the rule graph. Experimental results on four real-world datasets demonstrate that AnoT surpasses existing methods significantly in terms of accuracy and interoperability. All of the raw datasets and the implementation of AnoT are provided in https://github.com/zjs123/ANoT.

In-Batch contrastive learning is a state-of-the-art self-supervised method that brings semantically-similar instances close while pushing dissimilar instances apart within a mini-batch. Its key to success is the negative sharing strategy, in which every instance serves as a negative for the others within the mini-batch. Recent studies aim to improve performance by sampling hard negatives within the current mini-batch, whose quality is bounded by the mini-batch itself. In this work, we propose to improve contrastive learning by sampling mini-batches from the input data. We present BatchSampler\footnoteThe code is available at BatchSampler to sample mini-batches of hard-to-distinguish (i.e., hard and true negatives to each other) instances. To make each mini-batch have fewer false negatives, we design the proximity graph of randomly-selected instances. To form the mini-batch, we leverage random walk with restart on the proximity graph to help sample hard-to-distinguish instances. BatchSampler is a simple and general technique that can be directly plugged into existing contrastive learning models in vision, language, and graphs. Extensive experiments on datasets of three modalities show that BatchSampler can consistently improve the performance of powerful contrastive models, as shown by significant improvements of SimCLR on ImageNet-100, SimCSE on STS (language), and GraphCL and MVGRL on graph datasets.

Connectivity across documents often exhibits a hierarchical network structure. Hyperbolic Graph Neural Networks (HGNNs) have shown promise in preserving network hierarchy. However, they do not model the notion of topics, thus document representations lack semantic interpretability. On the other hand, a corpus of documents usually has high variability in degrees of topic specificity. For example, some documents contain general content (e.g., sports), while others focus on specific themes (e.g., basketball and swimming). Topic models indeed model latent topics for semantic interpretability, but most assume a flat topic structure and ignore such semantic hierarchy. Given these two challenges, we propose a Hyperbolic Graph Topic Modeling Network to integrate both network hierarchy across linked documents and semantic hierarchy within texts into a unified HGNN framework. Specifically, we construct a two-layer document graph. Intra- and cross-layer encoding captures network hierarchy. We design a topic tree for text decoding to preserve semantic hierarchy and learn interpretable topics. Supervised and unsupervised experiments verify the effectiveness of our model.

Subsurface simulations use computational models to predict the flow of fluids (e.g., oil, water, gas) through porous media. These simulations are pivotal in industrial applications such as petroleum production, where fast and accurate models are needed for high-stake decision making, for example, for well placement optimization and field development planning. Classical finite difference numerical simulators require massive computational resources to model large-scale real-world reservoirs. Alternatively, streamline simulators and data-driven surrogate models are computationally more efficient by relying on approximate physics models, however they are insufficient to model complex reservoir dynamics at scale.

Abusive behavior in online retail websites and communities threatens the experience of regular community members. Such behavior often takes place within a complex, dynamic, and large-scale network of users interacting with items. Detecting abuse is challenging due to the scarcity of labeled abuse instances and complexity of combining temporal and network patterns while operating at a massive scale. Previous approaches to dynamic graph modeling either do not scale, do not effectively generalize from a few labeled instances, or compromise performance for scalability. Here we present BiDyn, a general method to detect abusive behavior in dynamic bipartite networks at scale, while generalizing from limited training labels. BiDyn develops an efficient hybrid RNN-GNN architecture trained via a novel stacked ensemble training scheme. We also propose a novel pre-training framework for dynamic graphs that helps to achieve superior performance at scale. Our approach outperforms recent large-scale dynamic graph baselines in an abuse classification task by up to 14% AUROC while requiring 10x less memory per training batch in both open and proprietary datasets.

Graph Neural Networks (GNNs) are based on repeated aggregations of information from nodes' neighbors in a graph. However, because nodes share many neighbors, a naive implementation leads to repeated and inefficient aggregations and represents significant computational overhead. Here we propose Hierarchically Aggregated computation Graphs(HAGs), a new GNN representation technique that explicitly avoids redundancy by managing intermediate aggregation results hierarchically and eliminates repeated computations and unnecessary data transfers in GNN training and inference. HAGs perform the same computations and give the same models/accuracy as traditional GNNs, but in a much shorter time dueto optimized computations. To identify redundant computations,we introduce an accurate cost function and use a novel search algorithm to find optimized HAGs. Experiments show that the HAG representation significantly outperforms the standard GNN by increasing the end-to-end training throughput by up to 2.8× and reducing the aggregations and data transfers in GNN training byup to 6.3× and 5.6×, with only 0.1% memory overhead. Overall,our results represent an important advancement in speeding-up and scaling-up GNNs without any loss in model predictive performance.

Recent advancements in deep neural networks for graph-structured data have led to state-of-the-art performance on recommender system benchmarks. However, making these methods practical and scalable to web-scale recommendation tasks with billions of items and hundreds of millions of users remains an unsolved challenge. Here we describe a large-scale deep recommendation engine that we developed and deployed at Pinterest. We develop a data-efficient Graph Convolutional Network (GCN) algorithm, which combines efficient random walks and graph convolutions to generate embeddings of nodes (i.e., items) that incorporate both graph structure as well as node feature information. Compared to prior GCN approaches, we develop a novel method based on highly efficient random walks to structure the convolutions and design a novel training strategy that relies on harder-and-harder training examples to improve robustness and convergence of the model. We also develop an efficient MapReduce model inference algorithm to generate embeddings using a trained model. Overall, we can train on and embed graphs that are four orders of magnitude larger than typical GCN implementations. We show how GCN embeddings can be used to make high-quality recommendations in various settings at Pinterest, which has a massive underlying graph with 3 billion nodes representing pins and boards, and 17 billion edges. According to offline metrics, user studies, as well as A/B tests, our approach generates higher-quality recommendations than comparable deep learning based systems. To our knowledge, this is by far the largest application of deep graph embeddings to date and paves the way for a new generation of web-scale recommender systems based on graph convolutional architectures.

Dynamic time warping (DTW) is a widely used curve similarity measure. We present a simple and efficient (1 + ε)- approximation algorithm for DTW between a pair of point sequences, say, P and Q, each of which is sampled from a curve. We prove that the running time of the algorithm is O([EQUATION]n log σ) for a pair of k-packed curves with a total of n points, assuming that the spreads of P and Q are bounded by σ. The spread of a point set is the ratio of the maximum to the minimum pairwise distance, and a curve is called K- packed if the length of its intersection with any disk of radius r is at most Kr. Although an algorithm with similar asymptotic time complexity was presented in [1], our algorithm is considerably simpler and more efficient in practice.