Yuanchen Bei

Zihao Li, Kaifeng Jin, Yuanchen Bei et al.

Time series are often embedded in rich contexts that are essential for holistic modeling. Moreover, real-world practitioners often require end-to-end workflows for analyzing temporal dynamics, where widely studied tasks such as forecasting are only one step in a broader solution loop. While generalist AI agents offer a promising interface for such workflows under complex contexts, they still operate primarily in textual spaces that are not fully aligned with structured temporal signals. In this work, we introduce TimeClaw, an agentic harness framework for time series that equips generalist LLM agents with the time series-native runtime support needed for contextualized temporal reasoning. TimeClaw integrates executable temporal tools for grounded and auditable analysis, experience-driven capability evolution for creating reusable analytical routines, and episodic multimodal memory for retrieving relevant reasoning traces. Together, these components unlock harnessed open-ended temporal reasoning with contextual information. Extensive evaluation on multiple benchmarks covering diverse tasks across energy, finance, weather, traffic, and other real-world domains demonstrates improved performance of TimeClaw. Code is available at https://github.com/iDEA-iSAIL-Lab-UIUC/TimeClaw.

Yanjun Zhao, Tianxin Wei, Jiaru Zou et al.

Understanding scientific papers requires more than answering isolated questions or summarizing content. It involves an integrated reasoning process that grounds textual and visual information, interprets experimental evidence, synthesizes information across sources, and critically evaluates scientific claims. However, existing benchmarks typically assess these abilities in isolation, making it difficult to evaluate scientific paper understanding as a unified set of interacting cognitive abilities. In this work, we introduce PAPERMIND, a benchmark designed to evaluate integrated and agent-oriented scientific reasoning over research papers. PAPERMIND is constructed from real scientific papers across seven domains, including agriculture, biology, chemistry, computer science, medicine, physics, and economics. It comprises four complementary task families that collectively operationalize distinct cognitive facets of scientific paper reasoning, including multimodal grounding, experimental interpretation, cross-source evidence reasoning, and critical assessment. By analyzing model behavior across multiple tasks, PAPERMIND enables a diagnostic evaluation of integrated scientific reasoning behaviors that are difficult to assess through isolated task evaluations. Extensive experiments on both opensource and closed-source multimodal LLMs reveal consistent performance gaps across tasks, highlighting persistent challenges in integrated scientific reasoning and critique. Our benchmark and dataset are available at https:// github.com/Yanjun-Zhao/PaperMind.

Yuanchen Bei, Hao Xu, Sheng Zhou et al.

Dynamic graph data mining has gained popularity in recent years due to the rich information contained in dynamic graphs and their widespread use in the real world. Despite the advances in dynamic graph neural networks (DGNNs), the rich information and diverse downstream tasks have posed significant difficulties for the practical application of DGNNs in industrial scenarios. To this end, in this paper, we propose to address them by pre-training and present the Contrastive Pre-Training Method for Dynamic Graph Neural Networks (CPDG). CPDG tackles the challenges of pre-training for DGNNs, including generalization capability and long-short term modeling capability, through a flexible structural-temporal subgraph sampler along with structural-temporal contrastive pre-training schemes. Extensive experiments conducted on both large-scale research and industrial dynamic graph datasets show that CPDG outperforms existing methods in dynamic graph pre-training for various downstream tasks under three transfer settings.

Yue Xu, Hao Chen, Zengde Deng et al.

Graph Convolutional Networks (GCNs) and their variants have achieved significant performances on various recommendation tasks. However, many existing GCN models tend to perform recursive aggregations among all related nodes, which can arise severe computational burden to hinder their application to large-scale recommendation tasks. To this end, this paper proposes the flattened GCN~(FlatGCN) model, which is able to achieve superior performance with remarkably less complexity compared with existing models. Our main contribution is three-fold. First, we propose a simplified but powerful GCN architecture which aggregates the neighborhood information using one flattened GCN layer, instead of recursively. The aggregation step in FlatGCN is parameter-free such that it can be pre-computed with parallel computation to save memory and computational cost. Second, we propose an informative neighbor-infomax sampling method to select the most valuable neighbors by measuring the correlation among neighboring nodes based on a principled metric. Third, we propose a layer ensemble technique which improves the expressiveness of the learned representations by assembling the layer-wise neighborhood representations at the final layer. Extensive experiments on three datasets verify that our proposed model outperforms existing GCN models considerably and yields up to a few orders of magnitude speedup in training efficiency.

Qinggang Zhang, Shengyuan Chen, Yuanchen Bei et al.

Large language models (LLMs) have demonstrated remarkable capabilities in a wide range of tasks, yet their application to specialized domains remains challenging due to the need for deep expertise. Retrieval-Augmented generation (RAG) has emerged as a promising solution to customize LLMs for professional fields by seamlessly integrating external knowledge bases, enabling real-time access to domain-specific expertise during inference. Despite its potential, traditional RAG systems, based on flat text retrieval, face three critical challenges: (i) complex query understanding in professional contexts, (ii) difficulties in knowledge integration across distributed sources, and (iii) system efficiency bottlenecks at scale. This survey presents a systematic analysis of Graph-based Retrieval-Augmented Generation (GraphRAG), a new paradigm that revolutionizes domain-specific LLM applications. GraphRAG addresses traditional RAG limitations through three key innovations: (i) graph-structured knowledge representation that explicitly captures entity relationships and domain hierarchies, (ii) efficient graph-based retrieval techniques that enable context-preserving knowledge retrieval with multihop reasoning ability, and (iii) structure-aware knowledge integration algorithms that leverage retrieved knowledge for accurate and logical coherent generation of LLMs. In this survey, we systematically analyze the technical foundations of GraphRAG and examine current implementations across various professional domains, identifying key technical challenges and promising research directions. All the related resources of GraphRAG, including research papers, open-source data, and projects, are collected for the community in https://github.com/DEEP-PolyU/Awesome-GraphRAG.

Weizhi Zhang, Yuanchen Bei, Liangwei Yang et al. · tsinghua

Cold-start problem is one of the long-standing challenges in recommender systems, focusing on accurately modeling new or interaction-limited users or items to provide better recommendations. Due to the diversification of internet platforms and the exponential growth of users and items, the importance of cold-start recommendation (CSR) is becoming increasingly evident. At the same time, large language models (LLMs) have achieved tremendous success and possess strong capabilities in modeling user and item information, providing new potential for cold-start recommendations. However, the research community on CSR still lacks a comprehensive review and reflection in this field. Based on this, in this paper, we stand in the context of the era of large language models and provide a comprehensive review and discussion on the roadmap, related literature, and future directions of CSR. Specifically, we have conducted an exploration of the development path of how existing CSR utilizes information, from content features, graph relations, and domain information, to the world knowledge possessed by large language models, aiming to provide new insights for both the research and industrial communities on CSR. Related resources of cold-start recommendations are collected and continuously updated for the community in https://github.com/YuanchenBei/Awesome-Cold-Start-Recommendation.

Yijie Zhang, Yuanchen Bei, Shiqi Yang et al.

Graph collaborative filtering, which learns user and item representations through message propagation over the user-item interaction graph, has been shown to effectively enhance recommendation performance. However, most current graph collaborative filtering models mainly construct the interaction graph on a single behavior domain (e.g. click), even though users exhibit various types of behaviors on real-world platforms, including actions like click, cart, and purchase. Furthermore, due to variations in user engagement, there exists an imbalance in the scale of different types of behaviors. For instance, users may click and view multiple items but only make selective purchases from a small subset of them. How to alleviate the behavior imbalance problem and utilize information from the multiple behavior graphs concurrently to improve the target behavior conversion (e.g. purchase) remains underexplored. To this end, we propose IMGCF, a simple but effective model to alleviate behavior data imbalance for multi-behavior graph collaborative filtering. Specifically, IMGCF utilizes a multi-task learning framework for collaborative filtering on multi-behavior graphs. Then, to mitigate the data imbalance issue, IMGCF improves representation learning on the sparse behavior by leveraging representations learned from the behavior domain with abundant data volumes. Experiments on two widely-used multi-behavior datasets demonstrate the effectiveness of IMGCF.

Xuying Ning, Katherine Tieu, Dongqi Fu et al.

Recent large language models (LLMs) have demonstrated strong capabilities in understanding and generating code, from competitive programming to repository-level software engineering. In emerging agentic systems, code is no longer only a target output. It increasingly serves as an operational substrate for agent reasoning, acting, environment modeling, and execution-based verification. We frame this shift through the lens of agent harnesses and introduce code as agent harness: a unified view that centers code as the basis for agent infrastructure. To systematically study this perspective, we organize the survey around three connected layers. First, we study the harness interface, where code connects agents to reasoning, action, and environment modeling. Second, we examine harness mechanisms: planning, memory, and tool use for long-horizon execution, together with feedback-driven control and optimization that make harness reliable and adaptive. Third, we discuss scaling the harness from single-agent systems to multi-agent settings, where shared code artifacts support multi-agent coordination, review, and verification. Across these layers, we summarize representative methods and practical applications of code as agent harness, spanning coding assistants, GUI/OS automation, embodied agents, scientific discovery, personalization and recommendation, DevOps, and enterprise workflows. We further outline open challenges for harness engineering, including evaluation beyond final task success, verification under incomplete feedback, regression-free harness improvement, consistent shared state across multiple agents, human oversight for safety-critical actions, and extensions to multimodal environments. By centering code as the harness of agentic AI, this survey provides a unified roadmap toward executable, verifiable, and stateful AI agent systems.

Weizhi Zhang, Yangning Li, Yuanchen Bei et al. · pku

Information retrieval is a cornerstone of modern knowledge acquisition, enabling billions of queries each day across diverse domains. However, traditional keyword-based search engines are increasingly inadequate for handling complex, multi-step information needs. Our position is that Large Language Models (LLMs), endowed with reasoning and agentic capabilities, are ushering in a new paradigm termed Agentic Deep Research. These systems transcend conventional information search techniques by tightly integrating autonomous reasoning, iterative retrieval, and information synthesis into a dynamic feedback loop. We trace the evolution from static web search to interactive, agent-based systems that plan, explore, and learn. We also introduce a test-time scaling law to formalize the impact of computational depth on reasoning and search. Supported by benchmark results and the rise of open-source implementations, we demonstrate that Agentic Deep Research not only significantly outperforms existing approaches, but is also poised to become the dominant paradigm for future information seeking. All the related resources, including industry products, research papers, benchmark datasets, and open-source implementations, are collected for the community in https://github.com/DavidZWZ/Awesome-Deep-Research.

Yuanchen Bei, Tianxin Wei, Xuying Ning et al.

Long-term memory is a critical capability for multimodal large language model (MLLM) agents, particularly in conversational settings where information accumulates and evolves over time. However, existing benchmarks either evaluate multi-session memory in text-only conversations or assess multimodal understanding within localized contexts, failing to evaluate how multimodal memory is preserved, organized, and evolved across long-term conversational trajectories. Thus, we introduce Mem-Gallery, a new benchmark for evaluating multimodal long-term conversational memory in MLLM agents. Mem-Gallery features high-quality multi-session conversations grounded in both visual and textual information, with long interaction horizons and rich multimodal dependencies. Building on this dataset, we propose a systematic evaluation framework that assesses key memory capabilities along three functional dimensions: memory extraction and test-time adaptation, memory reasoning, and memory knowledge management. Extensive benchmarking across thirteen memory systems reveals several key findings, highlighting the necessity of explicit multimodal information retention and memory organization, the persistent limitations in memory reasoning and knowledge management, as well as the efficiency bottleneck of current models.

Yangning Li, Weizhi Zhang, Yuyao Yang et al. · pku

Retrieval-Augmented Generation (RAG) lifts the factuality of Large Language Models (LLMs) by injecting external knowledge, yet it falls short on problems that demand multi-step inference; conversely, purely reasoning-oriented approaches often hallucinate or mis-ground facts. This survey synthesizes both strands under a unified reasoning-retrieval perspective. We first map how advanced reasoning optimizes each stage of RAG (Reasoning-Enhanced RAG). Then, we show how retrieved knowledge of different type supply missing premises and expand context for complex inference (RAG-Enhanced Reasoning). Finally, we spotlight emerging Synergized RAG-Reasoning frameworks, where (agentic) LLMs iteratively interleave search and reasoning to achieve state-of-the-art performance across knowledge-intensive benchmarks. We categorize methods, datasets, and open challenges, and outline research avenues toward deeper RAG-Reasoning systems that are more effective, multimodally-adaptive, trustworthy, and human-centric. The collection is available at https://github.com/DavidZWZ/Awesome-RAG-Reasoning.

Guanbao Liang, Yuanchen Bei, Sheng Zhou et al.

Automatic prompt optimization is a promising approach for adapting large language models (LLMs) to downstream tasks, yet existing methods typically search for a specific prompt specialized to a fixed task. This paradigm limits generalization across heterogeneous queries and prevents models from accumulating reusable prompting knowledge over time. In this paper, we propose MemAPO, a memory-driven framework that reconceptualizes prompt optimization as generalizable and self-evolving experience accumulation. MemAPO maintains a dual-memory mechanism that distills successful reasoning trajectories into reusable strategy templates while organizing incorrect generations into structured error patterns that capture recurrent failure modes. Given a new prompt, the framework retrieves both relevant strategies and failure patterns to compose prompts that promote effective reasoning while discouraging known mistakes. Through iterative self-reflection and memory editing, MemAPO continuously updates its memory, enabling prompt optimization to improve over time rather than restarting from scratch for each task. Experiments on diverse benchmarks show that MemAPO consistently outperforms representative prompt optimization baselines while substantially reducing optimization cost.

Chengyu Lai, Sheng Zhou, Zhimeng Jiang et al.

Recommendation systems play a pivotal role in suggesting items to users based on their preferences. However, in online platforms, these systems inevitably offer unsuitable recommendations due to limited model capacity, poor data quality, or evolving user interests. Enhancing user experience necessitates efficiently rectify such unsuitable recommendation behaviors. This paper introduces a novel and significant task termed recommendation editing, which focuses on modifying known and unsuitable recommendation behaviors. Specifically, this task aims to adjust the recommendation model to eliminate known unsuitable items without accessing training data or retraining the model. We formally define the problem of recommendation editing with three primary objectives: strict rectification, collaborative rectification, and concentrated rectification. Three evaluation metrics are developed to quantitatively assess the achievement of each objective. We present a straightforward yet effective benchmark for recommendation editing using novel Editing Bayesian Personalized Ranking Loss. To demonstrate the effectiveness of the proposed method, we establish a comprehensive benchmark that incorporates various methods from related fields. Codebase is available at https://github.com/cycl2018/Recommendation-Editing.

Yuanchen Bei, Sheng Zhou, Qiaoyu Tan et al.

Unsupervised graph anomaly detection is crucial for various practical applications as it aims to identify anomalies in a graph that exhibit rare patterns deviating significantly from the majority of nodes. Recent advancements have utilized Graph Neural Networks (GNNs) to learn high-quality node representations for anomaly detection by aggregating information from neighborhoods. However, the presence of anomalies may render the observed neighborhood unreliable and result in misleading information aggregation for node representation learning. Selecting the proper neighborhood is critical for graph anomaly detection but also challenging due to the absence of anomaly-oriented guidance and the interdependence with representation learning. To address these issues, we utilize the advantages of reinforcement learning in adaptively learning in complex environments and propose a novel method that incorporates Reinforcement neighborhood selection for unsupervised graph ANomaly Detection (RAND). RAND begins by enriching the candidate neighbor pool of the given central node with multiple types of indirect neighbors. Next, RAND designs a tailored reinforcement anomaly evaluation module to assess the reliability and reward of considering the given neighbor. Finally, RAND selects the most reliable subset of neighbors based on these rewards and introduces an anomaly-aware aggregator to amplify messages from reliable neighbors while diminishing messages from unreliable ones. Extensive experiments on both three synthetic and two real-world datasets demonstrate that RAND outperforms the state-of-the-art methods.

Junnan Dong, Zijin Hong, Yuanchen Bei et al.

Large language models (LLMs) have demonstrated their remarkable performance across various language understanding tasks. While emerging benchmarks have been proposed to evaluate LLMs in various domains such as mathematics and computer science, they merely measure the accuracy in terms of the final prediction on multi-choice questions. However, it remains insufficient to verify the essential understanding of LLMs given a chosen choice. To fill this gap, we present CLR-Bench to comprehensively evaluate the LLMs in complex college-level reasoning. Specifically, (i) we prioritize 16 challenging college disciplines in computer science and artificial intelligence. The dataset contains 5 types of questions, while each question is associated with detailed explanations from experts. (ii) To quantify a fair evaluation of LLMs' reasoning ability, we formalize the criteria with two novel metrics. Q$\rightarrow$A is utilized to measure the performance of direct answer prediction, and Q$\rightarrow$AR effectively considers the joint ability to answer the question and provide rationale simultaneously. Extensive experiments are conducted with 40 LLMs over 1,018 discipline-specific questions. The results demonstrate the key insights that LLMs, even the best closed-source LLM, i.e., GPT-4 turbo, tend to `guess' the college-level answers. It shows a dramatic decrease in accuracy from 63.31% Q$\rightarrow$A to 39.00% Q$\rightarrow$AR, indicating an unsatisfactory reasoning ability.

Yuanchen Bei, Weizhi Zhang, Siwen Wang et al.

AI agents have experienced a paradigm shift, from early dominance by reinforcement learning (RL) to the rise of agents powered by large language models (LLMs), and now further advancing towards a synergistic fusion of RL and LLM capabilities. This progression has endowed AI agents with increasingly strong abilities. Despite these advances, to accomplish complex real-world tasks, agents are required to plan and execute effectively, maintain reliable memory, and coordinate smoothly with other agents. Achieving these capabilities involves contending with ever-present intricate information, operations, and interactions. In light of this challenge, data structurization can play a promising role by transforming intricate and disorganized data into well-structured forms that agents can more effectively understand and process. In this context, graphs, with their natural advantage in organizing, managing, and harnessing intricate data relationships, present a powerful data paradigm for structurization to support the capabilities demanded by advanced AI agents. To this end, this survey presents a first systematic review of how graphs can empower AI agents. Specifically, we explore the integration of graph techniques with core agent functionalities, highlight notable applications, and identify prospective avenues for future research. By comprehensively surveying this burgeoning intersection, we hope to inspire the development of next-generation AI agents equipped to tackle increasingly sophisticated challenges with graphs. Related resources are collected and continuously updated for the community in the Github link.

Yuanchen Bei, Weizhi Chen, Hao Chen et al.

Multi-label node classification is an important yet under-explored domain in graph mining as many real-world nodes belong to multiple categories rather than just a single one. Although a few efforts have been made by utilizing Graph Convolution Networks (GCNs) to learn node representations and model correlations between multiple labels in the embedding space, they still suffer from the ambiguous feature and ambiguous topology induced by multiple labels, which reduces the credibility of the messages delivered in graphs and overlooks the label correlations on graph data. Therefore, it is crucial to reduce the ambiguity and empower the GCNs for accurate classification. However, this is quite challenging due to the requirement of retaining the distinctiveness of each label while fully harnessing the correlation between labels simultaneously. To address these issues, in this paper, we propose a Correlation-aware Graph Convolutional Network (CorGCN) for multi-label node classification. By introducing a novel Correlation-Aware Graph Decomposition module, CorGCN can learn a graph that contains rich label-correlated information for each label. It then employs a Correlation-Enhanced Graph Convolution to model the relationships between labels during message passing to further bolster the classification process. Extensive experiments on five datasets demonstrate the effectiveness of our proposed CorGCN.

Yuanchen Bei, Sheng Zhou, Jinke Shi et al.

Unsupervised graph anomaly detection aims at identifying rare patterns that deviate from the majority in a graph without the aid of labels, which is important for a variety of real-world applications. Recent advances have utilized Graph Neural Networks (GNNs) to learn effective node representations by aggregating information from neighborhoods. This is motivated by the hypothesis that nodes in the graph tend to exhibit consistent behaviors with their neighborhoods. However, such consistency can be disrupted by graph anomalies in multiple ways. Most existing methods directly employ GNNs to learn representations, disregarding the negative impact of graph anomalies on GNNs, resulting in sub-optimal node representations and anomaly detection performance. While a few recent approaches have redesigned GNNs for graph anomaly detection under semi-supervised label guidance, how to address the adverse effects of graph anomalies on GNNs in unsupervised scenarios and learn effective representations for anomaly detection are still under-explored. To bridge this gap, in this paper, we propose a simple yet effective framework for Guarding Graph Neural Networks for Unsupervised Graph Anomaly Detection (G3AD). Specifically, G3AD first introduces two auxiliary networks along with correlation constraints to guard the GNNs against inconsistent information encoding. Furthermore, G3AD introduces an adaptive caching module to guard the GNNs from directly reconstructing the observed graph data that contains anomalies. Extensive experiments demonstrate that our G3AD can outperform twenty state-of-the-art methods on both synthetic and real-world graph anomaly datasets, with flexible generalization ability in different GNN backbones.

Tianxin Wei, Noveen Sachdeva, Benjamin Coleman et al.

Statefulness is essential for large language model (LLM) agents to perform long-term planning and problem-solving. This makes memory a critical component, yet its management and evolution remain largely underexplored. Existing evaluations mostly focus on static conversational settings, where memory is passively retrieved from dialogue to answer queries, overlooking the dynamic ability to accumulate and reuse experience across evolving task streams. In real-world environments such as interactive problem assistants or embodied agents, LLMs are required to handle continuous task streams, yet often fail to learn from accumulated interactions, losing valuable contextual insights, a limitation that calls for test-time evolution, where LLMs retrieve, integrate, and update memory continuously during deployment. To bridge this gap, we introduce Evo-Memory, a comprehensive streaming benchmark and framework for evaluating self-evolving memory in LLM agents. Evo-Memory structures datasets into sequential task streams, requiring LLMs to search, adapt, and evolve memory after each interaction. We unify and implement over ten representative memory modules and evaluate them across 10 diverse multi-turn goal-oriented and single-turn reasoning and QA datasets. To better benchmark experience reuse, we provide a baseline method, ExpRAG, for retrieving and utilizing prior experience, and further propose ReMem, an action-think-memory refine pipeline that tightly integrates reasoning, task actions, and memory updates to achieve continual improvement.