Coding agents powered by large language models (LLMs) have gained traction for automating code generation through iterative problem-solving with minimal human involvement. Despite the emergence of various frameworks, e.g., LangChain, AutoML, and AIDE, ML scientists still struggle to effectively review and adjust the agents' coding process. The current approach of manually inspecting individual outputs is inefficient, making it difficult to track code evolution, compare coding iterations, and identify improvement opportunities. To address this challenge, we introduce a visual analytics system designed to enhance the examination of coding agent behaviors. Focusing on the AIDE framework, our system supports comparative analysis across three levels: (1) Code-Level Analysis, which reveals how the agent debugs and refines its code over iterations; (2) Process-Level Analysis, which contrasts different solution-seeking processes explored by the agent; and (3) LLM-Level Analysis, which highlights variations in coding behavior across different LLMs. By integrating these perspectives, our system enables ML scientists to gain a structured understanding of agent behaviors, facilitating more effective debugging and prompt engineering. Through case studies using coding agents to tackle popular Kaggle competitions, we demonstrate how our system provides valuable insights into the iterative coding process.

As large language models (LLMs) are deployed in safety-critical settings, it is essential to ensure that their responses comply with safety standards. Prior research has revealed that LLMs often fail to grasp the notion of safe behaviors, resulting in either unjustified refusals to harmless prompts or the generation of harmful content. While substantial efforts have been made to improve their robustness, existing defenses often rely on costly fine-tuning of model parameters or employ suboptimal heuristic techniques. In this work, we take a novel approach to safeguard LLMs by learning to adapt the system prompts in instruction-tuned LLMs. While LLMs are typically pre-trained to follow a fixed system prompt, we investigate the impact of tailoring the system prompt to each specific user input on the safety of the responses. To this end, we propose Sysformer, a transformer model that updates an initial system prompt to a more robust system prompt in the LLM input embedding space while attending to the user prompt. While keeping the LLM parameters frozen, the Sysformer is trained to refuse to respond to a set of harmful prompts while responding ideally to a set of safe ones. Through extensive experiments on 5 LLMs from different families and 2 recent benchmarks, we demonstrate that Sysformer can significantly enhance the robustness of LLMs, leading to upto 80% gain in the refusal rate on harmful prompts while enhancing the compliance with the safe prompts by upto 90%. Results also generalize well to sophisticated jailbreaking attacks, making LLMs upto 100% more robust against different attack strategies. We hope our findings lead to cheaper safeguarding of LLMs and motivate future investigations into designing variable system prompts.

Retrieval-augmented Generation (RAG) extends large language models (LLMs) with external knowledge but faces key challenges: restricted effective context length and redundancy in retrieved documents. Pure compression-based approaches reduce input size but often discard fine-grained details essential for factual accuracy. We propose SARA, a unified RAG framework that balances local precision and global knowledge coverage under tight context budgets. SARA combines natural-language text snippets with semantic compression vectors to jointly enhance context efficiency and answer correctness. It represents contexts at two complementary levels: 1) fine-grained natural-language spans that preserve critical entities and numerical values, and 2) compact, interpretable vectors that summarize high-level semantics. An iterative evidence-selection module employs the compression vectors for dynamic reranking of contexts. Across 9 datasets and 5 open-source LLMs spanning 3 model families (Mistral, Llama, and Gemma), SARA consistently improves answer relevance (+17.71), answer correctness (+13.72), and semantic similarity (+15.53), demonstrating the importance of integrating textual and compressed representations for robust, context-efficient RAG.

Graph Neural Networks (GNNs) combine node attributes over a fixed granularity of the local graph structure around a node to predict its label. However, different nodes may relate to a node-level property with a different granularity of its local neighborhood, and using the same level of smoothing for all nodes can be detrimental to their classification. In this work, we challenge the common fact that a single GNN layer can classify all nodes of a graph by training GNNs with a distinct personalized layer for each node. Inspired by metric learning, we propose a novel algorithm, MetSelect1, to select the optimal representation layer to classify each node. In particular, we identify a prototype representation of each class in a transformed GNN layer and then, classify using the layer where the distance is smallest to a class prototype after normalizing with that layer's variance. Results on 10 datasets and 3 different GNNs show that we significantly improve the node classification accuracy of GNNs in a plug-and-play manner. We also find that using variable layers for prediction enables GNNs to be deeper and more robust to poisoning attacks. We hope this work can inspire future works to learn more adaptive and personalized graph representations.

Rationale discovery is defined as finding a subset of the input data that maximally supports the prediction of downstream tasks. In graph machine learning context, graph rationale is defined to locate the critical subgraph in the given graph topology, which fundamentally determines the prediction results. In contrast to the rationale subgraph, the remaining subgraph is named the environment subgraph. Graph rationalization can enhance the model performance as the mapping between the graph rationale and prediction label is viewed as invariant, by assumption. To ensure the discriminative power of the extracted rationale subgraphs, a key technique named "intervention" is applied. The core idea of intervention is that given any changing environment subgraphs, the semantics from the rationale subgraph is invariant, which guarantees the correct prediction result. However, most, if not all, of the existing rationalization works on graph data develop their intervention strategies on the graph level, which is coarse-grained. In this paper, we propose well-tailored intervention strategies on graph data. Our idea is driven by the development of Transformer models, whose self-attention module provides rich interactions between input nodes. Based on the self-attention module, our proposed invariant graph Transformer (IGT) can achieve fine-grained, more specifically, node-level and virtual node-level intervention. Our comprehensive experiments involve 7 real-world datasets, and the proposed IGT shows significant performance advantages compared to 13 baseline methods.

Large Language Models (LLMs) are becoming essential tools for various natural language processing tasks but often suffer from generating outdated or incorrect information. Retrieval-Augmented Generation (RAG) addresses this issue by incorporating external, real-time information retrieval to ground LLM responses. However, the existing RAG systems frequently struggle with the quality of retrieval documents, as irrelevant or noisy documents degrade performance, increase computational overhead, and undermine response reliability. To tackle this problem, we propose Multi-Agent Filtering Retrieval-Augmented Generation (MAIN-RAG), a training-free RAG framework that leverages multiple LLM agents to collaboratively filter and score retrieved documents. Specifically, MAIN-RAG introduces an adaptive filtering mechanism that dynamically adjusts the relevance filtering threshold based on score distributions, effectively minimizing noise while maintaining high recall of relevant documents. The proposed approach leverages inter-agent consensus to ensure robust document selection without requiring additional training data or fine-tuning. Experimental results across four QA benchmarks demonstrate that MAIN-RAG consistently outperforms traditional RAG approaches, achieving a 2-11 improvement in answer accuracy while reducing the number of irrelevant retrieved documents. Quantitative analysis further reveals that our approach achieves superior response consistency and answer accuracy over baseline methods, offering a competitive and practical alternative to training-based solutions.

Most existing personalized federated learning approaches are based on intricate designs, which often require complex implementation and tuning. In order to address this limitation, we propose a simple yet effective personalized federated learning framework. Specifically, during each communication round, we group clients into multiple clusters based on their model training status and data distribution on the server side. We then consider each cluster center as a node equipped with model parameters and construct a graph that connects these nodes using weighted edges. Additionally, we update the model parameters at each node by propagating information across the entire graph. Subsequently, we design a precise personalized model distribution strategy to allow clients to obtain the most suitable model from the server side. We conduct experiments on three image benchmark datasets and create synthetic structured datasets with three types of typologies. Experimental results demonstrate the effectiveness of the proposed FedCedar.

Contrastive Learning (CL) has shown promising performance in collaborative filtering. The key idea is to generate augmentation-invariant embeddings by maximizing the Mutual Information between different augmented views of the same instance. However, we empirically observe that existing CL models suffer from the dimensional collapse issue, where user/item embeddings only span a low-dimension subspace of the entire feature space. This suppresses other dimensional information and weakens the distinguishability of embeddings. Here we propose a non-contrastive learning objective, named nCL, which explicitly mitigates dimensional collapse of representations in collaborative filtering. Our nCL aims to achieve geometric properties of Alignment and Compactness on the embedding space. In particular, the alignment tries to push together representations of positive-related user-item pairs, while compactness tends to find the optimal coding length of user/item embeddings, subject to a given distortion. More importantly, our nCL does not require data augmentation nor negative sampling during training, making it scalable to large datasets. Experimental results demonstrate the superiority of our nCL.

Network embedding plays an important role in a variety of social network applications. Existing network embedding methods, explicitly or implicitly, can be categorized into positional embedding (PE) methods or structural embedding (SE) methods. Specifically, PE methods encode the positional information and obtain similar embeddings for adjacent/close nodes, while SE methods aim to learn identical representations for nodes with the same local structural patterns, even if the two nodes are far away from each other. The disparate designs of the two types of methods lead to an apparent dilemma in that no embedding could perfectly capture both positional and structural information. In this paper, we seek to demystify the underlying relationship between positional embedding and structural embedding. We first point out that the positional embedding can produce the structural embedding with simple transformations, while the opposite direction cannot hold. Based on this finding, a novel network embedding model PACER is proposed, which optimizes the positional embedding with the help of random walk with restart (RWR) proximity distribution, and such positional embedding is then used to seamlessly obtain the structural embedding with simple transformations. Furthermore, two variants of PACER are proposed to handle node classification task on homophilic and heterophilic graphs. Extensive experiments on 17 datasets show that PACER achieves comparable or better performance than the state-of-the-arts.

Bipartite graphs are ubiquitous in real-world applications, where butterflies (2x2-bicliques) serve as fundamental building blocks for analyzing graph motifs. While existing wedge aggregation based butterfly computation methods are effective for small-scale graphs, they face significant scalability challenges when processing large-scale bipartite graphs with billions of vertices and edges. In this paper, we present MONARCH, a distributed framework for efficient butterfly computation. Unlike traditional approaches that require expensive two-hop neighbor traversal, MONARCH operates exclusively on first-hop neighbor information, reducing both computation time and communication overhead. This pioneering framework expands the boundary of butterfly computation methods, making it feasible to process bipartite graphs that are orders of magnitude larger. To demonstrate its scalability, we show experimental results on bipartite graph with billions of vertices and edges.

Graph anomaly detection (GAD) aims to find network elements (e.g., nodes, edges) with significantly atypical patterns and has a profound impact in a variety of application domains, including social network analysis, security, Web, finance, and many more. Most of the existing methods have been developed in an unsupervised manner or with extremely limited supervision, due to the high cost of acquiring ground-truth information. Consequently, the identified anomalies may turn out to be noises or uneventful instances because of the lack of prior knowledge on graph anomalies. To address the data scarcity issue in GAD, in this paper, we propose, gADAM, a novel graph neural network-based GAD framework, which consolidates (1) an innovative mixup approach to augment the original training data by adaptively interpolating data instances in the embedding space, and (2) an efficacious sampling method to obtain high-quality negative samples for model training. Additionally, to advance the representation learning for GAD, we further equip the proposed framework with a generic prototype-based learning module. Through extensive empirical evaluations, we corroborate the superiority of the proposed gADAM framework on graph anomaly detection w.r.t. various metrics.

Graph is a prevalent discrete data structure, whose generation has wide applications such as drug discovery and circuit design. Diffusion generative models, as an emerging research focus, have been applied to graph generation tasks. Overall, according to the space of states and time steps, diffusion generative models can be categorized into discrete-/continuous-state discrete-/continuous-time fashions. In this paper, we formulate the graph diffusion generation in a discrete-state continuous-time setting, which has never been studied in previous graph diffusion models. The rationale of such a formulation is to preserve the discrete nature of graph-structured data and meanwhile provide flexible sampling trade-offs between sample quality and efficiency. Analysis shows that our training objective is closely related to the generation quality and our proposed generation framework enjoys ideal invariant/equivariant properties concerning the permutation of node ordering. Our proposed model shows competitive empirical performance against other state-of-the-art graph generation solutions on various benchmarks while at the same time can flexibly trade off the generation quality and efficiency in the sampling phase.

Approximately repeated subsequences in a longer time series, i.e., time series motifs, are important primitive in time series data mining. Motifs are used in dozens of downstream tasks, including classification, clustering, summarization, rule discovery, segmentation etc. Time series motif discovery is a notoriously computationally expensive task. Some motif discovery algorithms are fast in the best case, but in other datasets, even if both the data and motif lengths are held the same, both their time and space complexity can explode. The Matrix Profile has the nice property that its time and space complexity are independent of the data. Moreover, the Matrix Profile is fast enough for datasets with about one million datapoints, which covers a large fraction of user cases. However, there are situations where we may wish to consider datasets which are much larger. In this work, we introduce the first lower bound for the Matrix Profile and an algorithm that exploits that lower bound to allow orders of magnitude speed up for exact motif search on real-world datasets. We demonstrate the utility of our ideas with the largest and most ambitious motif discovery experiments ever attempted.

Knowledge Graph Embedding (KGE) is a powerful technique for predicting missing links in Knowledge Graphs (KGs) by learning the entities and relations. Hyperbolic space has emerged as a promising embedding space for KGs due to its ability to represent hierarchical data. Nevertheless, most existing hyperbolic KGE methods rely on tangent approximation and are not fully hyperbolic, resulting in distortions and inaccuracies. To overcome this limitation, we propose LorentzKG, a fully hyperbolic KGE method that represents entities as points in the Lorentz model and represents relations as the intrinsic transformation-the Lorentz transformations between entities. We demonstrate that the Lorentz transformation, which can be decomposed into Lorentz rotation/reflection and Lorentz boost, captures various types of relations including hierarchical structures. Experimental results show that our LorentzKG achieves state-of-the-art performance(1).

We consider the problem of test-time adaptation of predictive models trained on tabular data. Effective solution of this problem requires adaptation of predictive models trained on the source domain to a target domain, using only unlabeled target domain data, without access to source domain data. Existing test-time adaptation methods for tabular data have difficulty coping with the heterogeneous features and their complex dependencies inherent in tabular data. To overcome these limitations, we consider test-time adaptation in the setting wherein the logical structure of the rules is assumed to remain invariant despite distribution shift between source and target domains whereas the numerical parameters associated with the rules and the weights assigned to them can vary to accommodate distribution shift. TabLog discretizes numerical features, models dependencies between heterogeneous features, introduces a novel contrastive loss for coping with distribution shift, and presents an end-to-end framework for efficient training and test-time adaptation by taking advantage of a logical neural network representation of a rule ensemble. We present results of experiments using several benchmark data sets that demonstrate TabLog is competitive with or improves upon the state-of-the-art methods for test-time adaptation of predictive models trained on tabular data. Our code is available at https://github.com/WeijieyingRen/TabLog.

Knowledge graph question answering aims to identify answers of the query according to the facts in the knowledge graph. In the vast majority of the existing works, the input queries are considered perfect and can precisely express the user’s query intention. However, in reality, input queries might be ambiguous and elusive which only contain a limited amount of information. Directly answering these ambiguous queries may yield unwanted answers and deteriorate user experience. In this paper, we propose PReFNet which focuses on answering ambiguous queries with pseudo relevance feedback on knowledge graphs. In order to leverage the hidden (pseudo) relevance information existed in the results that are initially returned from a given query, PReFNet treats the top-k returned candidate answers as a set of most relevant answers, and uses variational Bayesian inference to infer user’s query intention. To boost the quality of the inferred queries, a neighborhood embedding based VGAE model is used to prune inferior inferred queries. The inferred high quality queries will be returned to the users to help them search with ease. Moreover, all the high-quality candidate nodes will be re-ranked according to the inferred queries. The experiment results show that our proposed method can recommend high-quality query graphs to users and improve the question answering accuracy.

Graph Neural Networks (GNNs) have achieved impressive performance in collaborative filtering. However, recent studies show that GNNs tend to yield inferior performance when the distributions of training and test data are not aligned well. Moreover, training GNNs often requires optimizing non-convex neural networks with an abundance of local and global minima, which may differ widely in their performance at test time. Thus, it is essential to develop an optimization strategy that can choose the minima carefully, which can yield strong generalization performance on unseen data. Here we propose an effective training schema, called gSAM, under the principle that theflatter minima has a better generalization ability than thesharper ones. To achieve this goal, gSAM regularizes the flatness of the weight loss landscape by forming a bi-level optimization: the outer problem conducts the standard model training while the inner problem helps the model jump out of the sharp minima. Experimental results show the superiority of our gSAM.

Finding the proper depth $d$ of a GNN that provides strong representation power has drawn significant attention, yet nonetheless largely remains an open problem for the graph learning community. Although noteworthy progress has been made, the depth or the number of layers of a corresponding GCN is realized by a series of graph convolution operations, which naturally makes $d$ a positive integer ($d \in \mathbb{N}+$). An interesting question is whether breaking the constraint of $\mathbb{N}+$ by making $d$ a real number ($d \in \mathbb{R}$) can bring new insights into graph learning mechanisms. In this work, by redefining GCN's depth $d$ as a trainable parameter continuously adjustable within $(-\infty,+\infty)$, we open a new door of controlling its expressiveness on graph signal processing to model graph homophily/heterophily (nodes with similar/dissimilar labels/attributes tend to inter-connect). A simple and powerful GCN model ReD-GCN, is proposed to retain the simplicity of GCN and meanwhile automatically search for the optimal $d$ without the prior knowledge regarding whether the input graph is homophilic or heterophilic. Negative-valued $d$ intrinsically enables high-pass frequency filtering functionality for graph heterophily. Variants extending the model flexibility/scalability are also developed. The theoretical feasibility of having a real-valued depth with explainable physical meanings is ensured via eigen-decomposition of the graph Laplacian and a properly designed transformation function from the perspective of functional calculus. Extensive experiments demonstrate the superiority of ReD-GCN on node classification tasks for a variety of graphs. Furthermore, by introducing the concept of eigengraph, a novel graph augmentation method is obtained: the optimal $d$ effectively generates a new topology through a properly weighted combination of eigengraphs, which dramatically boosts the performance even for a vanilla GCN.