Pith. sign in

REVIEW 4 major objections 4 minor 3 cited by

Graph-Augmented Large Language Model Agents: Current Progress and Future Prospects

T0 review · 4 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Graphs are the missing infrastructure for reliable, long-horizon, and coordinated LLM agents, according to a survey that maps the field with a six-way functional taxonomy.

desk verdict A useful first roadmap for graph-augmented LLM agents, with a functional taxonomy that is more pragmatic than perfect; the 'comprehensive' claim needs a documented selection protocol and the self-citation bias disclosed. read the letter →

arxiv 2507.21407 v2 pith:JF56VO5S submitted 2025-07-29 cs.AI

Yixin Liu , Guibin Zhang , Kun Wang , Shiyuan Li , Shirui Pan This is my paper
classification cs.AI
keywords Graph-augmentedLLMagentsmulti-agentsystemsgraphneuralnetworksplanningmemorytoolmanagementorchestrationtrustworthiness
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Large language models alone make unreliable agents: they hallucinate plans, forget across long interactions, struggle to select from large tool sets, and coordinate poorly in groups. This survey argues that graphs—tool graphs, plan graphs, knowledge graphs, and agent coordination graphs—are a general fix that supplies structure, continuity, and coordination. It organizes the fast-growing and fragmented research area into a functional taxonomy: planning, memory, and tool usage for single agents, plus orchestration, efficiency, and trustworthiness for multi-agent systems. If the taxonomy holds, it gives researchers a shared vocabulary and a roadmap for where graph methods plug into agent systems and where the open problems lie.

What carries the argument

The central object is the graph as an auxiliary structure in agent systems: nodes are sub-tasks, memories, tools, or agents, and edges encode dependencies, semantic relations, input-output compatibility, or communication channels. The paper's load-bearing mechanism is the functional taxonomy built on this object, which assigns each GLA method a primary function—planning, memory, or tool usage for single agents, and orchestration, efficiency, or trustworthiness for multi-agent systems. The taxonomy is what turns scattered papers into a roadmap, and the graph structures are what the roadmap is about.

What would settle it

A systematic literature search with pre-registered inclusion criteria that found a substantial share of GLA methods cannot be assigned a single primary function—or that found an important class of methods missing from all six categories—would falsify the taxonomy's completeness and mutual exclusivity.

Watch

Extended reading notes

Core claim

The paper's central claim is that graphs are not a niche add-on but a cross-cutting infrastructure for LLM agents. In the single-agent case, graphs make planning reliable by representing plans, sub-task pools, reasoning traces, and environments as structured graphs; they make memory persistent by organizing interaction histories and knowledge bases as graphs; and they make tool use scalable by encoding tools as a navigable graph. In the multi-agent case, graphs are the natural language for orchestration, moving from static topologies to task-adaptive and process-adaptive ones, and graph-theoretic ideas of edge, node, and layer redundancy explain when communication should be pruned. The paper further argues that graph structure supports trustworthiness by modeling how threats propagate through agent networks. It presents these categories as the organizing framework for the field and points future work toward dynamic graph learning, unified full-stack graph abstractions, multimodal graphs, and large-scale simulation.

Load-bearing premise

The survey's load-bearing premise is that its six-way functional taxonomy faithfully and completely partitions the graph-augmented agent design space, so that the selected papers are representative and every method has one primary role.

Editorial extensions

If this is right

  • Graph-based planning methods can replace or supervise LLM planners, reducing hallucinated and non-executable plans by grounding sub-tasks in a constrained task graph.
  • Graph-organized memory lets agents retain and retrieve long interaction histories and external knowledge without relying on the LLM's context window.
  • Tool graphs turn tool selection into graph search or navigation, so agents can handle tool sets far larger than an LLM can reason over directly.
  • Pruning redundant agents and communication edges—analogies of graph sparsification and dropout—can cut token cost without losing performance.
  • If the roadmap is adopted, new GLA work can be positioned by which module it augments and which graph abstraction it uses, making results easier to compare.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The taxonomy's three-plus-three split implies that methods can be classified by primary function, but several surveyed systems, such as workflow graphs that serve both planning and orchestration, show the categories are not strictly mutually exclusive; a composite label may eventually be needed.
  • The survey's efficiency section suggests a transferable design principle: techniques from graph neural network training, such as dropout and residual connections, may apply to agent teams as generic recipes rather than one-off fixes.
  • A testable extension would be to use the taxonomy to predict research gaps: categories with few representative methods, such as process-dynamic topology or trustworthy multi-agent systems, would be the places where new contributions are most likely to matter.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. The paper surveys Graph-augmented LLM Agents (GLA), defining GLA as LLM-based agents whose modules or execution flows are organized, enhanced, or interpreted using graphs. It proposes a functional taxonomy: for single-agent systems, graphs for planning (plan-as-graph, sub-task pool, reasoning thought, environment), memory (interaction and knowledge memory), and tool management (selection and tool-use training); for multi-agent systems, graphs for orchestration (static, task-dynamic, and process-dynamic topologies), efficiency (edge, node, and layer redundancy), and trustworthiness. It concludes with five future directions: dynamic and continual graph learning, unified graph abstractions, multimodal graphs, graphs for trustworthy MAS, and large-scale MAS simulation. The authors claim that this is the first comprehensive taxonomy and review article on GLA.

Significance. The paper addresses a genuine and timely topic: the use of graphs to structure planning, memory, tool use, and multi-agent coordination in LLM agents. Its functional taxonomy is intuitive, and the figures (especially Figures 3, 4, 6, and 7) effectively communicate the different roles graphs can play. The efficiency section, which draws an analogy between MAS redundancy and graph-structure phenomena such as edge redundancy, node redundancy, and oversmoothing, is a distinctive and useful angle. The future-directions section is thought-provoking and likely to be useful to new researchers. However, the paper's central claim of being a comprehensive overview is not yet fully supported: the selection of works is not shown to be systematic or representative, the taxonomy categories overlap in places, and several strong novelty claims are unverified. These issues do not require new experiments; they can be addressed by adding a methodology statement, broadening or clarifying the scope, and softening the claims.

major comments (4)
  1. [Abstract; Section 1] The paper's central claim is that 'this paper offers a timely and comprehensive overview of recent advances' in GLA, but the manuscript provides no systematic literature search protocol, no inclusion/exclusion criteria, and no validation of the survey's coverage. The linked GitHub repository is mentioned but not audited, and the reference list is not justified as representative. As a result, the comprehensiveness claim is under-supported. Please add a methodology subsection describing the search process (databases, time window, keywords, screening) and any inter-annotator validation of the taxonomy; alternatively, explicitly reframe the paper as a curated, non-exhaustive survey.
  2. [Section 1, paragraph 4] The statement 'there has been, to date, no comprehensive taxonomy or review article on this line of work' is asserted without citation or search evidence, and similar unverified 'first' claims appear in Section 3.1.2 (G-Designer makes the first attempt), Section 3.2.1 (AgentPrune is the first to formally define communication redundancy), and Section 3.2.3 (this phenomenon was first observed). These strong novelty claims are load-bearing for the paper's motivation and should either be substantiated with a targeted literature check or softened.
  3. [Sections 2.1.1, 3.1.1, 3.1.2, 3.2, 3.3] The selection of representative methods shows a visible concentration of the authors' own works (e.g., G-Designer, MaAS, EvoFlow, AgentPrune, G-Safeguard, NetSafe, AgentSafe, ARG-Designer, and EIB-learner), while widely used graph-based agent frameworks such as LangGraph and graph-RAG-style knowledge memory systems are absent. This pattern weakens the claim of a balanced comprehensive overview. Please add a coverage analysis or an explicit acknowledgment of selection constraints, and consider discussing prominent missing frameworks.
  4. [Sections 2.1.1, 3.1.1, 3.1.3] The taxonomy is not a clean partition of the GLA design space: AFlow is presented as 'plan as a graph' in Section 2.1.1 and again as static MAS orchestration in Section 3.1.1; AgentKit's dynamic execution graph is discussed only under planning even though it is also a workflow graph; and ReSo, classified as process-dynamic orchestration in Section 3.1.3, is fundamentally a DAG plan. Because the roadmap's value depends on the categories being meaningful and non-overlapping, please clarify the classification rule (e.g., primary function versus implementation) and explicitly handle cross-listed methods.
minor comments (4)
  1. [Section 2.1.3; Figure 3] There is a duplicated phrase in Section 2.1.3 ('...biological processesin uncovering semantic relationships between biological processes'), and Figure 3 contains a typo ('Froutend' should be 'Frontend'). Please proofread the manuscript.
  2. [References] Some references are bibliographically incomplete: SciToolAgent (Chen et al., 2025b) lacks an arXiv identifier or venue, and Manish (2024) and Anthropic (2024) are not fully specified. Please complete these entries.
  3. [Sections 2.1.2, 2.2.2, 3.2.1] Several passages report quantitative superiority or benchmark results without sufficient context or critical appraisal, for example Section 2.1.2 ('Empirical evidence shows... outperforming LLM-based planners'), Section 2.2.2 ('outperforming larger models'), and Section 3.2.1 ('many systems can achieve comparable performance even after pruning'). As a survey aiming to guide the community, please add task, metric, and baseline details for these claims, or mark them as the cited authors' findings rather than established consensus.
  4. [Section 4.5] The claim that 'most existing graph-based MAS approaches are confined to small-scale scenarios, typically modeling only a few dozen agents' is not supported by a citation or measurement. Please provide evidence or qualify the statement.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular derivation: the paper is a survey whose taxonomy does not reduce to its inputs; minor self-citation is present but not load-bearing.

full rationale

The manuscript makes no empirical derivation or prediction: it is a taxonomy/overview of graph-augmented LLM agents. Its central claim is that the proposed functional organization (planning, memory, tool use for single agents; orchestration, efficiency, trustworthiness for multi-agent systems) is a comprehensive and useful roadmap. I checked whether any category or narrative is defined in terms of its own conclusion. It is not: each category is populated by multiple independently authored systems (e.g., GPTSwarm, MacNet, ControlLLM, ToT/GoT, AutoGen, ReSo, EvoMAC, AgentDropout, TalkHier, FlowReasoner) alongside some current-author works (AgentPrune, G-Designer, EIB-learner, ARG-Designer, MaAS, EvoFlow, NetSafe, AgentSafe, G-Safeguard, AgentAuditor). No equation is fitted and renamed as a prediction; no uniqueness theorem is imported from the authors' prior work; no ansatz is smuggled in via citation to justify the taxonomy. The absence of a documented search protocol and the observed overlap between categories (e.g., AFlow as both plan graph and static MAS workflow) are coverage/validity concerns for the 'comprehensive overview' claim, not circularity. The self-citation cluster in Sections 3.1.2 and 3.3 is noticeable but the surveyed claims do not reduce to those citations, so it does not raise the circularity score beyond a minor 2.

Assumptions & free parameters 0 free parameters · 2 assumptions · 0 invented entities

No free parameters or invented entities are introduced because the paper is a survey. The main load-bearing construct is the taxonomy itself, which is an ad hoc organizational schema. The survey also relies on the domain assumption that graphs mitigate LLM agent limitations.

assumptions (2)
  • domain assumption Large language models are insufficient for reliable planning, long-term memory, and tool management, so auxiliary graph structures provide a meaningful improvement.
    This premise motivates the entire survey; it is supported by the cited works but not independently established in this paper (Section 1).
  • ad hoc to paper The proposed taxonomy (planning, memory, tool usage; orchestration, efficiency, trustworthiness) is a natural and exhaustive partition of the GLA field.
    This categorization is introduced by the authors as an organizing principle, without a systematic derivation or comparison to alternative taxonomies such as by graph type or by learning algorithm.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Graph-Augmented Large Language Model Agents: Current Progress and Future Prospects." pith.science (2026). https://pith.science/paper/JF56VO5S

@misc{pith2026250721407,
  author       = {Pith},
  title        = {Pith review of: Graph-Augmented Large Language Model Agents: Current Progress and Future Prospects},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JF56VO5S}},
  note         = {Machine review of arXiv:2507.21407}
}
read the original abstract

Autonomous agents based on large language models (LLMs) have demonstrated impressive capabilities in a wide range of applications, including web navigation, software development, and embodied control. While most LLMs are limited in several key agentic procedures, such as reliable planning, long-term memory, tool management, and multi-agent coordination, graphs can serve as a powerful auxiliary structure to enhance structure, continuity, and coordination in complex agent workflows. Given the rapid growth and fragmentation of research on Graph-augmented LLM Agents (GLA), this paper offers a timely and comprehensive overview of recent advances and also highlights key directions for future work. Specifically, we categorize existing GLA methods by their primary functions in LLM agent systems, including planning, memory, and tool usage, and then analyze how graphs and graph learning algorithms contribute to each. For multi-agent systems, we further discuss how GLA solutions facilitate the orchestration, efficiency optimization, and trustworthiness of MAS. Finally, we highlight key future directions to advance this field, from improving structural adaptability to enabling unified, scalable, and multimodal GLA systems. We hope this paper can serve as a roadmap for future research on GLA and foster a deeper understanding of the role of graphs in LLM agent systems.

Figures

Figures reproduced from arXiv: 2507.21407 by the authors.

Figure 1
Figure 1. LLM agent framework and multi-agent sys [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Different graphs in LLM agent systems. tent reasoning over unfamiliar or similar tools (Liu et al., 2024b). Moreover, when we extend a single agent to multi-agent systems, how to manage inter￾agent communication and coordination remains an open question for LLMs (Guo et al., 2024). To address these limitations, researchers have in￾creasingly turned to graphs as a complementary in￾frastructure to organize, enhance, a… view at source ↗
Figure 3
Figure 3. Graphs for planning in LLM agent system. [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Graphs for memory management in LLM agent system. evant contexts, reduce hallucination, and signifi￾cantly improve bug localization accuracy. 2.2 Graphs for Agent Memory Management Effective memory management is essential for LLM agents to operate in complex, multi-ste…
Figure 5
Figure 5. Figure 5: Graphs for tool management in LLM agent system. we can not only perform accurate selection and retrieval of tools, but also enhance the tool-using capabilities of LLM agents. 2.3.1 Tool graphs for tool selection As the number of available tools grows, select￾ing the ap…
Figure 6
Figure 6. Figure 6: Graphs for MAS Orchestration. properly. To enhance the tool-use capability of LLM agents, supervised fine-tuning with high￾quality tool-interaction data has become an effec￾tive solution. To construct tool-interaction data, tool graphs offer a practical solution to sam…
Figure 7
Figure 7. Figure 7: Graphs for MAS Efficiency. topologies affect MAS performance. 3.1.3 Process-dynamic MAS Topology While these methods enable task-level adaptivity, they suffer from a key limitation: topologies are statically sampled prior to execution, lacking fine￾grained adaptation a…

Discussion (0). Sign in to comment.

Forward citations

Cited by 3 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Towards Self-Evolving Agents: A Human-Inspired Adaptive Exploration-Exploitation Framework for Genetic Network Programming

    cs.NE 2026-07 conditional novelty 6.0 of 10

    HGNP improves GNP and its variants via adaptive crossover protecting high-in-degree nodes later, early-favoring mutation of judgment-to-judgment links, and cycle elimination, with HGNP-SBGNP best on Tileworld.

  2. Toward Auditable Fraud Detection: Combining Graph Features, Model Explanations, and Agentic Case Investigation

    cs.LG 2026-07 conditional novelty 4.0 of 10

    On PaySim, graph and autoencoder signals improve fraud ranking only among mid-confidence cases, and a bounded LLM agent underperforms direct classifier thresholding.

  3. Securing Multi-Agent Systems Against Corruptions via Node Contribution Backpropagation

    cs.CR 2025-10 conditional novelty 4.0 of 10

    A backward-propagation scoring scheme over a signed temporal DAG can identify malicious agents in LLM multi-agent systems and cut their communications, improving defended accuracy by 3–7 percentage points in the autho...

Reference graph

Works this paper leans on

95 extracted references · 38 canonical work pages · cited by 3 Pith papers

  1. [1]

    Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, and 1 others. 2023. GPT-4 technical report. arXiv preprint arXiv:2303.08774

    arXiv 2023

  2. [2]

    Petr Anokhin, Nikita Semenov, Artyom Sorokin, Dmitry Evseev, Mikhail Burtsev, and Evgeny Burnaev. 2024. Arigraph: Learning knowledge graph world models with episodic memory for llm agents. arXiv preprint arXiv:2407.04363

    arXiv 2024

  3. [3]

    Anthropic. 2024. https://www.anthropic.com/engineering/building-effective-agents Building effective ai agents

    2024

  4. [4]

    Dario Badagliacca, Gabriele Caruso, Agnese Augello, and Luca Sabatucci. 2025. Graph of goal-oriented thoughts: Design and implementation of llm agents. In SOCIALIZE 2025, CEUR Workshop Proceedings

    2025

  5. [5]

    Maciej Besta, Nils Blach, Ales Kubicek, Robert Gerstenberger, Michal Podstawski, Lukas Gianinazzi, Joanna Gajda, Tomasz Lehmann, Hubert Niewiadomski, Piotr Nyczyk, and 1 others. 2024. Graph of thoughts: Solving elaborate problems with large language models. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 17682--17690

    2024

  6. [6]

    Chen Cai and Yusu Wang. 2020. A note on over-smoothing for graph neural networks. In ICML 2020 Workshop on Graph Representation Learning and Beyond

    2020

  7. [7]

    Chi-Min Chan, Weize Chen, Yusheng Su, Jianxuan Yu, Wei Xue, Shanghang Zhang, Jie Fu, and Zhiyuan Liu. 2023. Chateval: Towards better llm-based evaluators through multi-agent debate. In Twelfth International Conference on Learning Representations

    2023

  8. [8]

    Ada Chen, Yongjiang Wu, Junyuan Zhang, Shu Yang, Jen-tse Huang, Kun Wang, Wenxuan Wang, and Shuai Wang. 2025 a . A survey on the safety and security threats of computer-using agents: Jarvis or ultron? arXiv preprint arXiv:2505.10924

    arXiv 2025

Show all 95 references
  1. [9]

    Deli Chen, Yankai Lin, Wei Li, Peng Li, Jie Zhou, and Xu Sun. 2020. Measuring and relieving the over-smoothing problem for graph neural networks from the topological view. In Proceedings of the AAAI conference on artificial intelligence, volume 34, pages 3438--3445

    2020

  2. [10]

    Huaben Chen, Wenkang Ji, Lufeng Xu, and Shiyu Zhao. 2023. Multi-agent consensus seeking via large language models. arXiv preprint arXiv:2310.20151

    2023 arXiv

  3. [11]

    Huajun Chen, Keyan Ding, Jing Yu, Junjie Huang, Yuchen Yang, and Qiang Zhang. 2025 b . Scitoolagent: A knowledge graph-driven scientific agent for multi-tool integration

    2025

  4. [12]

    Zhaoling Chen, Xiangru Tang, Gangda Deng, Fang Wu, Jialong Wu, Zhiwei Jiang, Viktor Prasanna, Arman Cohan, and Xingyao Wang. 2025 c . Locagent: Graph-guided llm agents for code localization. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics

    2025

  5. [13]

    Yilun Du, Shuang Li, Antonio Torralba, Joshua B Tenenbaum, and Igor Mordatch. 2024. Improving factuality and reasoning in language models through multiagent debate. In Proceedings of the 41st International Conference on Machine Learning, pages 11733--11763

    2024

  6. [14]

    Sugyeong Eo, Hyeonseok Moon, Evelyn Hayoon Zi, Chanjun Park, and Heuiseok Lim. 2025. Debate only when necessary: Adaptive multiagent collaboration for efficient llm reasoning. arXiv preprint arXiv:2504.05047

    2025 arXiv

  7. [15]

    Wenqi Fan, Yujuan Ding, Liangbo Ning, Shijie Wang, Hengyun Li, Dawei Yin, Tat-Seng Chua, and Qing Li. 2024. A survey on rag meeting llms: Towards retrieval-augmented large language models. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining,...

    2024

  8. [16]

    Hongcheng Gao, Yue Liu, Yufei He, Longxu Dou, Chao Du, Zhijie Deng, Bryan Hooi, Min Lin, and Tianyu Pang. 2025. Flowreasoner: Reinforcing query-level meta-agents. arXiv preprint arXiv:2504.15257

    2025 arXiv

  9. [17]

    Taicheng Guo, Xiuying Chen, Yaqi Wang, Ruidi Chang, Shichao Pei, Nitesh V Chawla, Olaf Wiest, and Xiangliang Zhang. 2024. Large language model based multi-agents: a survey of progress and challenges. In Proceedings of the Thirty-Third International Joint Conference on Artifici...

    2024

  10. [18]

    Chi-Yang Hsu, Kyle Cox, Jiawei Xu, Zhen Tan, Tianhua Zhai, Mengzhou Hu, Dexter Pratt, Tianlong Chen, Ziniu Hu, and Ying Ding. 2024. Thought graph: Generating thought process for biological reasoning. In Companion Proceedings of the ACM Web Conference 2024, pages 537--540

    2024

  11. [19]

    Yue Hu, Yuzhu Cai, Yaxin Du, Xinyu Zhu, Xiangrui Liu, Zijie Yu, Yuchen Hou, Shuo Tang, and Siheng Chen. 2025. Self-evolving multi-agent collaboration networks for software development. In Thirteenth International Conference on Learning Representations

    2025

  12. [20]

    Wanjing Huang, Tongjie Pan, and Yalan Ye. 2025. Graphormer-guided task planning: Beyond static rules with llm safety perception. arXiv preprint arXiv:2503.06866

    2025 arXiv

  13. [21]

    Xu Huang, Weiwen Liu, Xiaolong Chen, Xingmei Wang, Hao Wang, Defu Lian, Yasheng Wang, Ruiming Tang, and Enhong Chen. 2024. Understanding the planning of llm agents: A survey. arXiv preprint arXiv:2402.02716

    2024 arXiv

  14. [22]

    Jinhao Jiang, Kun Zhou, Wayne Xin Zhao, Yang Song, Chen Zhu, Hengshu Zhu, and Ji-Rong Wen. 2025. Kg-agent: An efficient autonomous agent framework for complex reasoning over knowledge graph. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics

    2025

  15. [23]

    Wei Jin, Yao Ma, Xiaorui Liu, Xianfeng Tang, Suhang Wang, and Jiliang Tang. 2020. Graph structure learning for robust graph neural networks. In Proceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining, pages 66--74

    2020

  16. [24]

    Akbir Khan, John Hughes, Dan Valentine, Laura Ruis, Kshitij Sachan, Ansh Radhakrishnan, Edward Grefenstette, Samuel R Bowman, Tim Rockt \"a schel, and Ethan Perez. 2024. Debating with more persuasive llms leads to more truthful answers. In Proceedings of the 41st International...

    2024

  17. [25]

    Shiyuan Li, Yixin Liu, Qingsong Wen, Chengqi Zhang, and Shirui Pan. 2025 a . Assemble your crew: Automatic multi-agent communication topology design via autoregressive graph generation. arXiv preprint arXiv:2507.18224

    2025

  18. [26]

    Yuhan Li, Zhixun Li, Peisong Wang, Jia Li, Xiangguo Sun, Hong Cheng, and Jeffrey Xu Yu. 2024 a . A survey of graph meets large language model: Progress and future directions. In Proceedings of the Thirty-Third International Joint Conference on Artificial Intelligence

    2024

  19. [27]

    Yunxuan Li, Yibing Du, Jiageng Zhang, Le Hou, Peter Grabowski, Yeqing Li, and Eugene Ie. 2024 b . Improving multi-agent debate with sparse communication topology. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 7281--7294

    2024

  20. [28]

    Zherui Li, Yan Mi, Zhenhong Zhou, Houcheng Jiang, Guibin Zhang, Kun Wang, and Junfeng Fang. 2025 b . Goal-aware identification and rectification of misinformation in multi-agent systems. arXiv preprint arXiv:2506.00509

    2025 arXiv

  21. [29]

    Zhixun Li, Dingshuo Chen, Tong Zhao, Daixin Wang, Hongrui Liu, Zhiqiang Zhang, Jun Zhou, and Jeffrey Xu Yu. 2025 c . Iceberg: Debiased self-training for class-imbalanced node classification. In Proceedings of the ACM on Web Conference 2025, pages 3160--3170

    2025

  22. [30]

    Zhixun Li, Liang Wang, Xin Sun, Yifan Luo, Yanqiao Zhu, Dingshuo Chen, Yingtao Luo, Xiangxin Zhou, Qiang Liu, Shu Wu, and 1 others. 2023. Gslb: The graph structure learning benchmark. Advances in Neural Information Processing Systems, 36:30306--30318

    2023

  23. [31]

    Tian Liang, Zhiwei He, Wenxiang Jiao, Xing Wang, Yan Wang, Rui Wang, Yujiu Yang, Shuming Shi, and Zhaopeng Tu. 2024. Encouraging divergent thinking in large language models through multi-agent debate. In Proceedings of the 2024 Conference on Empirical Methods in Natural Langua...

    2024

  24. [32]

    Aishan Liu, Zonghao Ying, Le Wang, Junjie Mu, Jinyang Guo, Jiakai Wang, Yuqing Ma, Siyuan Liang, Mingchuan Zhang, Xianglong Liu, and 1 others. 2025 a . Agentsafe: Benchmarking the safety of embodied agents on hazardous instructions. arXiv preprint arXiv:2506.14697

    2025

  25. [33]

    Aishan Liu, Yuguang Zhou, Xianglong Liu, Tianyuan Zhang, Siyuan Liang, Jiakai Wang, Yanjun Pu, Tianlin Li, Junqi Zhang, Wenbo Zhou, and 1 others. 2025 b . Compromising llm driven embodied agents with contextual backdoor attacks. IEEE Transactions on Information Forensics and Security

    2025

  26. [34]

    Aixin Liu, Bei Feng, Bing Xue, Bingxuan Wang, Bochao Wu, Chengda Lu, Chenggang Zhao, Chengqi Deng, Chenyu Zhang, Chong Ruan, and 1 others. 2024 a . DeepSeek-v3 technical report. arXiv preprint arXiv:2412.19437

    2024 arXiv

  27. [35]

    Xukun Liu, Zhiyuan Peng, Xiaoyuan Yi, Xing Xie, Lirong Xiang, Yuchen Liu, and Dongkuan Xu. 2024 b . Toolnet: Connecting large language models with massive tools via tool graph. arXiv preprint arXiv:2403.00839

    2024 arXiv

  28. [36]

    Zhaoyang Liu, Zeqiang Lai, Zhangwei Gao, Erfei Cui, Ziheng Li, Xizhou Zhu, Lewei Lu, Qifeng Chen, Yu Qiao, Jifeng Dai, and 1 others. 2024 c . Controlllm: Augment language models with tools by searching on graphs. In European Conference on Computer Vision, pages 89--105

    2024

  29. [37]

    Zijun Liu, Yanzhe Zhang, Peng Li, Yang Liu, and Diyi Yang. 2023. Dynamic llm-agent network: An llm-agent collaboration framework with agent team optimization. arXiv preprint arXiv:2310.02170

    2023 arXiv

  30. [38]

    Dongsheng Luo, Wei Cheng, Wenchao Yu, Bo Zong, Jingchao Ni, Haifeng Chen, and Xiang Zhang. 2021. Learning to drop: Robust graph neural network via topological denoising. In Proceedings of the 14th ACM international conference on web search and data mining, pages 779--787

    2021

  31. [39]

    Hanjun Luo, Shenyu Dai, Chiming Ni, Xinfeng Li, Guibin Zhang, Kun Wang, Tongliang Liu, and Hanan Salam. 2025 a . Agentauditor: Human-level safety and security evaluation for llm agents. arXiv preprint arXiv:2506.00641

    2025

  32. [40]

    Linhao Luo, Zicheng Zhao, Gholamreza Haffari, Dinh Phung, Chen Gong, and Shirui Pan. 2025 b . Gfm-rag: Graph foundation model for retrieval augmented generation. arXiv preprint arXiv:2502.01113

    2025

  33. [41]

    Bran, Sam Cox, Oliver Schilter, Carlo Baldassari, Andrew D White, and Philippe Schwaller

    Andres M. Bran, Sam Cox, Oliver Schilter, Carlo Baldassari, Andrew D White, and Philippe Schwaller. 2024. Augmenting large language models with chemistry tools. Nature Machine Intelligence, 6(5):525--535

    2024

  34. [42]

    Sanwal Manish. 2024. An autonomous multi-agent llm framework for agile software development. International Journal of Trend in Scientific Research and Development, 8(5):892--898

    2024

  35. [43]

    Junyuan Mao, Fanci Meng, Yifan Duan, Miao Yu, Xiaojun Jia, Junfeng Fang, Yuxuan Liang, Kun Wang, and Qingsong Wen. 2025. Agentsafe: Safeguarding large language model-based multi-agent systems via hierarchical data management. arXiv preprint arXiv:2503.04392

    2025 arXiv

  36. [44]

    Zepeng Ning and Lihua Xie. 2024. A survey on multi-agent reinforcement learning and its application. Journal of Automation and Intelligence, 3(2):73--91

    2024

  37. [45]

    P \'a l Andr \'a s Papp, Karolis Martinkus, Lukas Faber, and Roger Wattenhofer. 2021. Dropgnn: Random dropouts increase the expressiveness of graph neural networks. Advances in Neural Information Processing Systems, 34:21997--22009

    2021

  38. [46]

    Chen Qian, Zihao Xie, Yifei Wang, Wei Liu, Yufan Dang, Zhuoyun Du, Weize Chen, Cheng Yang, Zhiyuan Liu, and Maosong Sun. 2025. Scaling large-language-model-based multi-agent collaboration. In Thirteenth International Conference on Learning Representations

    2025

  39. [47]

    Shuofei Qiao, Runnan Fang, Zhisong Qiu, Xiaobin Wang, Ningyu Zhang, Yong Jiang, Pengjun Xie, Fei Huang, and Huajun Chen. 2025. Benchmarking agentic workflow generation. In Thirteenth International Conference on Learning Representations

    2025

  40. [48]

    Yu Rong, Wenbing Huang, Tingyang Xu, and Junzhou Huang. 2020. Dropedge: Towards deep graph convolutional networks on node classification. In Eight International Conference on Learning Representations

    2020

  41. [49]

    Xu Shen, Yixin Liu, Yiwei Dai, Yili Wang, Rui Miao, Yue Tan, Shirui Pan, and Xin Wang. 2025. Understanding the information propagation effects of communication topologies in llm-based multi-agent systems. In Findings of the Association for Computational Linguistics: EMNLP 2025

    2025

  42. [50]

    Yongliang Shen, Kaitao Song, Xu Tan, Dongsheng Li, Weiming Lu, and Yueting Zhuang. 2023. Hugginggpt: Solving ai tasks with chatgpt and its friends in hugging face. Advances in Neural Information Processing Systems, 36:38154--38180

    2023

  43. [51]

    Daniel A Spielman and Nikhil Srivastava. 2008. Graph sparsification by effective resistances. In Proceedings of the fortieth annual ACM symposium on Theory of computing, pages 563--568

    2008

  44. [52]

    Gemini Team, Petko Georgiev, Ving Ian Lei, Ryan Burnell, Libin Bai, Anmol Gulati, Garrett Tanzer, Damien Vincent, Zhufeng Pan, Shibo Wang, and 1 others. 2024. Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context. arXiv preprint arXiv:2403.05530

    2024 arXiv

  45. [53]

    Kun Wang, Guibin Zhang, Zhenhong Zhou, Jiahao Wu, Miao Yu, Shiqian Zhao, Chenlong Yin, Jinhu Fu, Yibo Yan, Hanjun Luo, and 1 others. 2025 a . A comprehensive survey in llm (-agent) full stack safety: Data, training and deployment. arXiv preprint arXiv:2504.15585

    2025 arXiv

  46. [54]

    Le Wang, Zonghao Ying, Tianyuan Zhang, Siyuan Liang, Shengshan Hu, Mingchuan Zhang, Aishan Liu, and Xianglong Liu. 2025 b . Manipulating multimodal agents via cross-modal prompt injection. arXiv preprint arXiv:2504.14348

    2025 arXiv

  47. [55]

    Lei Wang, Chen Ma, Xueyang Feng, Zeyu Zhang, Hao Yang, Jingsen Zhang, Zhiyuan Chen, Jiakai Tang, Xu Chen, Yankai Lin, and 1 others. 2024. A survey on large language model based autonomous agents. Frontiers of Computer Science, 18(6):186345

    2024

  48. [56]

    Shilong Wang, Guibin Zhang, Miao Yu, Guancheng Wan, Fanci Meng, Chongye Guo, Kun Wang, and Yang Wang. 2025 c . G-safeguard: A topology-guided security lens and treatment on llm-based multi-agent systems. In Proceedings of the 63rd Annual Meeting of the Association for Computat...

    2025

  49. [57]

    Song Wang, Zhen Tan, Zihan Chen, Shuang Zhou, Tianlong Chen, and Jundong Li. 2025 d . Anymac: Cascading flexible multi-agent collaboration via next-agent prediction. arXiv preprint arXiv:2506.17784

    2025

  50. [58]

    Zezhong Wang, Xingshan Zeng, Weiwen Liu, Liangyou Li, Yasheng Wang, Lifeng Shang, Xin Jiang, Qun Liu, and Kam-Fai Wong. 2025 e . Toolflow: Boosting llm tool-calling through natural and coherent dialogue synthesis. In Proceedings of the 2025 Conference of the Nations of the Ame...

    2025

  51. [59]

    Zhao Wang, Sota Moriyama, Wei-Yao Wang, Briti Gangopadhyay, and Shingo Takamatsu. 2025 f . Talk structurally, act hierarchically: A collaborative framework for llm multi-agent systems. arXiv preprint arXiv:2502.11098

    2025 arXiv

  52. [60]

    Zhexuan Wang, Yutong Wang, Xuebo Liu, Liang Ding, Miao Zhang, Jie Liu, and Min Zhang. 2025 g . Agentdropout: Dynamic agent elimination for token-efficient and high-performance llm-based multi-agent collaboration. arXiv preprint arXiv:2503.18891

    2025 arXiv

  53. [61]

    Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, and 1 others. 2022. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 35:24824--24837

    2022

  54. [62]

    Yangbo Wei, Zhen Huang, Huang Li, Wei W Xing, Ting-Jung Lin, and Lei He. 2025. Vflow: Discovering optimal agentic workflows for verilog generation. arXiv preprint arXiv:2504.03723

    2025 arXiv

  55. [63]

    Qingyun Wu, Gagan Bansal, Jieyu Zhang, Yiran Wu, Beibin Li, Erkang Zhu, Li Jiang, Xiaoyun Zhang, Shaokun Zhang, Jiale Liu, and 1 others. 2024 a . Autogen: Enabling next-gen llm applications via multi-agent conversation. In ICLR 2024 Workshop on Large Language Model (LLM) Agents

    2024

  56. [64]

    Xixi Wu, Yifei Shen, Caihua Shan, Kaitao Song, Siwei Wang, Bohang Zhang, Jiarui Feng, Hong Cheng, Wei Chen, Yun Xiong, and 1 others. 2024 b . Can graph learning improve planning in llm-based agents? In The Thirty-eighth Annual Conference on Neural Information Processing Systems

    2024

  57. [65]

    Yue Wu, Yewen Fan, So Yeon Min, Shrimai Prabhumoye, Stephen McAleer, Yonatan Bisk, Ruslan Salakhutdinov, Yuanzhi Li, and Tom Mitchell. 2024 c . Agentkit: Structured llm reasoning with dynamic graphs. In First Conference on Language Modeling

    2024

  58. [66]

    Zonghan Wu, Shirui Pan, Fengwen Chen, Guodong Long, Chengqi Zhang, and S Yu Philip. 2020. A comprehensive survey on graph neural networks. IEEE transactions on neural networks and learning systems, 32(1):4--24

    2020

  59. [67]

    Zhiheng Xi, Wenxiang Chen, Xin Guo, Wei He, Yiwen Ding, Boyang Hong, Ming Zhang, Junzhe Wang, Senjie Jin, Enyu Zhou, and 1 others. 2025. The rise and potential of large language model based agents: A survey. Science China Information Sciences, 68(2):121101

    2025

  60. [68]

    Zhentao Xie, Chengcheng Han, Jinxin Shi, Wenjun Cui, Xin Zhao, Xingjiao Wu, and Jiabao Zhao. 2025. Rmoa: Optimizing mixture-of-agents through diversity maximization and residual compensation. arXiv preprint arXiv:2505.24442

    2025 arXiv

  61. [69]

    Wujiang Xu, Kai Mei, Hang Gao, Juntao Tan, Zujie Liang, and Yongfeng Zhang. 2025. A-mem: Agentic memory for llm agents. arXiv preprint arXiv:2502.12110

    2025 arXiv

  62. [70]

    Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Tom Griffiths, Yuan Cao, and Karthik Narasimhan. 2023. Tree of thoughts: Deliberate problem solving with large language models. Advances in neural information processing systems, 36:11809--11822

    2023

  63. [71]

    Miao Yu, Fanci Meng, Xinyun Zhou, Shilong Wang, Junyuan Mao, Linsey Pang, Tianlong Chen, Kun Wang, Xinfeng Li, Yongfeng Zhang, and 1 others. 2025 a . A survey on trustworthy llm agents: Threats and countermeasures. In Proceedings of the 31th ACM SIGKDD international conference...

    2025

  64. [72]

    Miao Yu, Shilong Wang, Guibin Zhang, Junyuan Mao, Chenlong Yin, Qijiong Liu, Qingsong Wen, Kun Wang, and Yang Wang. 2025 b . Netsafe: Exploring the topological safety of multi-agent networks. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics

    2025

  65. [73]

    Yangyang Yu, Zhiyuan Yao, Haohang Li, Zhiyang Deng, Yuechen Jiang, Yupeng Cao, Zhi Chen, Jordan Suchow, Zhenyu Cui, Rong Liu, and 1 others. 2024. Fincon: A synthesized llm multi-agent system with conceptual verbal reinforcement for enhanced financial decision making. Advances ...

    2024

  66. [74]

    Simon Zhai, Hao Bai, Zipeng Lin, Jiayi Pan, Peter Tong, Yifei Zhou, Alane Suhr, Saining Xie, Yann LeCun, Yi Ma, and 1 others. 2024. Fine-tuning large vision-language models as decision-making agents via reinforcement learning. Advances in neural information processing systems,...

    2024

  67. [75]

    Chi Zhang, Zhao Yang, Jiaxuan Liu, Yanda Li, Yucheng Han, Xin Chen, Zebiao Huang, Bin Fu, and Gang Yu. 2025 a . Appagent: Multimodal agents as smartphone users. In Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems, pages 1--20

    2025

  68. [76]

    Guibin Zhang, Kaijie Chen, Guancheng Wan, Heng Chang, Hong Cheng, Kun Wang, Shuyue Hu, and Lei Bai. 2025 b . Evoflow: Evolving diverse agentic workflows on the fly. arXiv preprint arXiv:2502.07373

    2025 arXiv

  69. [77]

    Guibin Zhang, Luyang Niu, Junfeng Fang, Kun Wang, Lei Bai, and Xiang Wang. 2025 c . Multi-agent architecture search via agentic supernet. In Forty-second International Conference on Machine Learning

    2025

  70. [78]

    Guibin Zhang, Xiangguo Sun, Yanwei Yue, Chonghe Jiang, Kun Wang, Tianlong Chen, and Shirui Pan. 2025 d . Graph sparsification via mixture of graphs. In Thirteenth International Conference on Learning Representations

    2025

  71. [79]

    Guibin Zhang, Yanwei Yue, Zhixun Li, Sukwon Yun, Guancheng Wan, Kun Wang, Dawei Cheng, Jeffrey Xu Yu, and Tianlong Chen. 2025 e . Cut the crap: An economical communication pipeline for llm-based multi-agent systems. In Thirteenth International Conference on Learning Representations

    2025

  72. [80]

    Guibin Zhang, Yanwei Yue, Xiangguo Sun, Guancheng Wan, Miao Yu, Junfeng Fang, Kun Wang, Tianlong Chen, and Dawei Cheng. 2025 f . G-designer: Architecting multi-agent communication topologies via graph neural networks. In Forty-second International Conference on Machine Learning

    2025

  73. [81]

    Guibin Zhang, Yanwei Yue, Kun Wang, Junfeng Fang, Yongduo Sui, Kai Wang, Yuxuan Liang, Dawei Cheng, Shirui Pan, and Tianlong Chen. 2024 a . Two heads are better than one: boosting graph sparse training via semantic and topological awareness. In Proceedings of the 41st Internat...

    2024

  74. [82]

    Jiayi Zhang, Jinyu Xiang, Zhaoyang Yu, Fengwei Teng, Xionghui Chen, Jiaqi Chen, Mingchen Zhuge, Xin Cheng, Sirui Hong, Jinlin Wang, and 1 others. 2025 g . Aflow: Automating agentic workflow generation. In Thirteenth International Conference on Learning Representations

    2025

  75. [83]

    Jinghan Zhang, Xiting Wang, Weijieying Ren, Lu Jiang, Dongjie Wang, and Kunpeng Liu. 2025 h . Ratt: A thought structure for coherent and correct llm reasoning. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 39, pages 26733--26741

    2025

  76. [84]

    Jintian Zhang, Xin Xu, Ningyu Zhang, Ruibo Liu, Bryan Hooi, and Shumin Deng. 2024 b . Exploring collaboration mechanisms for llm agents: A social psychology view. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics

    2024

  77. [85]

    Shiqi Zhang, Xinbei Ma, Zouying Cao, Zhuosheng Zhang, and Hai Zhao. 2025 i . Plan-over-graph: Towards parallelable llm agent schedule. arXiv preprint arXiv:2502.14563

    2025 arXiv

  78. [86]

    Yuanshuo Zhang, Yuchen Hou, Bohan Tang, Shuo Chen, Muhan Zhang, Xiaowen Dong, and Siheng Chen. 2025 j . Gnns as predictors of agentic workflow performances. arXiv preprint arXiv:2503.11301

    2025 arXiv

  79. [87]

    Zhexin Zhang, Shiyao Cui, Yida Lu, Jingzhuo Zhou, Junxiao Yang, Hongning Wang, and Minlie Huang. 2024 c . Agent-safetybench: Evaluating the safety of llm agents. arXiv preprint arXiv:2412.14470

    2024 arXiv

  80. [88]

    Han Zhou, Xingchen Wan, Ruoxi Sun, Hamid Palangi, Shariq Iqbal, Ivan Vuli \'c , Anna Korhonen, and Sercan \"O Ar k. 2025 a . Multi-agent design: Optimizing agents with better prompts and topologies. arXiv preprint arXiv:2502.02533

    2025

  81. [89]

    Heng Zhou, Hejia Geng, Xiangyuan Xue, Li Kang, Yiran Qin, Zhiyong Wang, Zhenfei Yin, and Lei Bai. 2025 b . Reso: A reward-driven self-organizing llm-based multi-agent system for reasoning tasks. arXiv preprint arXiv:2503.02390

    2025 arXiv

  82. [90]

    Zhilun Zhou, Jingyang Fan, Yu Liu, Fengli Xu, Depeng Jin, and Yong Li. 2024. Synergizing llm agents and knowledge graph for socioeconomic prediction in lbsn. arXiv preprint arXiv:2411.00028

    2024 arXiv

  83. [91]

    Changxi Zhu, Mehdi Dastani, and Shihan Wang. 2024. A survey of multi-agent deep reinforcement learning with communication. Autonomous Agents and Multi-Agent Systems, 38(1):4

    2024

  84. [92]

    Yanqiao Zhu, Weizhi Xu, Jinghao Zhang, Yuanqi Du, Jieyu Zhang, Qiang Liu, Carl Yang, and Shu Wu. 2021. A survey on graph structure learning: Progress and opportunities. arXiv preprint arXiv:2103.03036

    2021 arXiv

  85. [93]

    Mingchen Zhuge, Wenyi Wang, Louis Kirsch, Francesco Faccio, Dmitrii Khizbullin, and J \"u rgen Schmidhuber. 2024. Gptswarm: Language agents as optimizable graphs. In Forty-first International Conference on Machine Learning

    2024

  86. [94]

    online" 'onlinestring :=

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list...

  87. [95]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.