arxiv: 2605.13172 · v1 · submitted 2026-05-13 · 💻 cs.MA · cs.AI

Recognition: 2 theorem links

· Lean Theorem

When Does Hierarchy Help? Benchmarking Agent Coordination in Event-Driven Industrial Scheduling

Hailiang Zhao, Wenzhuo Qian, Yuhao Yang, Zhiwei Ling, Ziqi Wang

Pith reviewed 2026-05-14 02:14 UTC · model grok-4.3

classification 💻 cs.MA cs.AI

keywords multi-agent coordinationevent-driven schedulinghierarchical agentsindustrial benchmarksagent robustnessconstraint satisfactionDESBenchcoordination paradigms

0 comments

The pith

Different coordination paradigms for agents in event-driven scheduling each carry distinct trade-offs in robustness, efficiency, alignment, and communication load.

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

The paper introduces DESBench, a benchmark built on a shared discrete-event environment, to test how multi-agent teams handle industrial scheduling tasks that involve coupled constraints, partial observability, and decisions at multiple time scales. It evaluates four representative coordination structures—centralized, hierarchical, heterarchical, and holonic—that differ in how information flows, who holds decision authority, and how conflicts are resolved. Controlled experiments measure not only task success but also constraint satisfaction, communication volume, and performance stability as problem difficulty increases. The results show clear patterns: centralized control stays reliable and communication-light yet cannot scale; hierarchical decomposition gains speed but produces level-to-level mismatches; heterarchical setups stay flexible yet require heavy messaging; holonic designs satisfy local constraints well yet lose overall robustness. These structural differences matter because most existing agent benchmarks only track final outcomes and therefore miss the coordination costs that determine whether a system can operate in real dynamic factories.

Core claim

Our controlled evaluations reveal clear coordination trade-offs: centralized coordination is robust and communication-efficient but scales poorly with difficulty; hierarchical coordination improves efficiency through decomposition but suffers from cross-level misalignment; heterarchical coordination is flexible but communication-heavy; and holonic coordination satisfies constraints well but loses global robustness.

What carries the argument

DESBench, a benchmark on a shared discrete-event driven environment that supplies tasks and metrics for effectiveness, constraint alignment, coordination efficiency, and robustness across four paradigms distinguished by their mechanisms of information flow, decision authority, and conflict resolution.

If this is right

Centralized coordination remains robust and communication-efficient yet cannot handle rising task difficulty.
Hierarchical coordination gains efficiency by decomposing problems but creates misalignment between decision levels.
Heterarchical coordination preserves flexibility at the price of higher communication demands.
Holonic coordination meets local constraints reliably yet fails to preserve overall system robustness.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Future agent systems could benefit from hybrid or adaptive coordination that switches structure according to current difficulty or observability level.
Benchmarks limited to outcome metrics will continue to overestimate the practical readiness of multi-agent systems for dynamic coupled environments.
Extending the same evaluation protocol to settings with stronger inter-agent coupling or longer time horizons would test whether the observed trade-off patterns persist.

Load-bearing premise

The four chosen coordination paradigms and the tasks and metrics defined in DESBench sufficiently capture the essential mechanisms of information flow, decision authority, and conflict resolution in real event-driven industrial systems that have partial observability.

What would settle it

Running identical scheduling scenarios on a physical industrial testbed that includes real sensor noise and communication delays and observing that centralized coordination scales without loss of robustness while holonic coordination retains global robustness would contradict the reported trade-offs.

Figures

Figures reproduced from arXiv: 2605.13172 by Hailiang Zhao, Wenzhuo Qian, Yuhao Yang, Zhiwei Ling, Ziqi Wang.

**Figure 2.** Figure 2: Overview of the DESBench. Agents at Plant, Area, and Cell levels receive local observations [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

read the original abstract

Recent advances in agent and multi-agent systems have shown strong performance on tool use, reasoning, and collaborative tasks. However, existing benchmarks mostly evaluate task completion in weakly coupled environments, and provide limited support for studying coordination in shared, dynamically evolving systems with hierarchy and coupled constraints. This leaves an important question underexplored: when do different coordination paradigms succeed or fail? We introduce Distributed Event-driven Scheduling Benchmark (DESBench), a benchmark for evaluating agent coordination in hierarchical event-driven scheduling. Built on a shared discrete-event driven environment in industrial scheduling, our benchmark captures multi-timescale decision making, partial observability, and dynamically coupled constraints. We define tasks and metrics that evaluate effectiveness, constraint alignment, coordination efficiency, and robustness, and focus on four representative coordination paradigms: centralized, hierarchical, heterarchical, and holonic. These paradigms correspond to distinct mechanisms of information flow, decision authority, and conflict resolution. Our controlled evaluations reveal clear coordination trade-offs: centralized coordination is robust and communication-efficient but scales poorly with difficulty; hierarchical coordination improves efficiency through decomposition but suffers from cross-level misalignment; heterarchical coordination is flexible but communication-heavy; and holonic coordination satisfies constraints well but loses global robustness. These findings demonstrate that coordination design fundamentally shapes agent system behavior in complex environments, revealing structural trade-offs that cannot be captured by outcome metrics alone and underscoring the imperative for more adaptive, principled, and dynamic coordination mechanisms in future MAS research.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

DESBench provides a useful new benchmark for coordination paradigms in industrial scheduling, though the experimental controls require close inspection.

read the letter

The punchline is that DESBench is a new benchmark for coordination in event-driven scheduling, and it compares four paradigms in a way that could help clarify when hierarchy helps or hurts in constrained settings. The paper does a good job identifying the limitations of existing benchmarks that focus on weakly coupled tasks. By building on a shared discrete-event environment with multi-timescale decision making and coupled constraints, it sets up tasks that better reflect industrial scheduling realities. The metrics for effectiveness, constraint alignment, coordination efficiency, and robustness are a reasonable way to evaluate the paradigms. What the paper does well is framing the distinct mechanisms: information flow, decision authority, and conflict resolution for each of centralized, hierarchical, heterarchical, and holonic coordination. The claimed trade-offs sound plausible on the surface. The soft spot is in the experimental validation. The abstract mentions controlled evaluations revealing specific trade-offs, but without seeing the details on how agent internals were standardized, it's possible the results are influenced by other factors like different planning algorithms or event generation. The stress-test concern about not isolating the coordination paradigm effects is a fair one to raise until the full protocol is examined. This paper is for colleagues working on multi-agent systems in applied domains like manufacturing or robotics. A reader who wants reproducible test cases for coordination mechanisms would get something from it. I recommend sending it to peer review. The benchmark itself is a contribution worth referee attention, and any issues with the controls can be addressed in revision.

Referee Report

1 major / 1 minor

Summary. The paper introduces DESBench, a benchmark for agent coordination in hierarchical event-driven industrial scheduling with partial observability and coupled constraints. It evaluates four paradigms (centralized, hierarchical, heterarchical, holonic) on metrics for effectiveness, constraint alignment, coordination efficiency, and robustness, claiming that controlled evaluations reveal specific trade-offs: centralized is robust and communication-efficient but scales poorly; hierarchical improves efficiency via decomposition but suffers cross-level misalignment; heterarchical is flexible but communication-heavy; and holonic satisfies constraints well but loses global robustness.

Significance. If the evaluations isolate coordination effects from other factors, the benchmark and trade-off findings would provide useful empirical guidance on coordination design in complex MAS for dynamic industrial settings, emphasizing structural properties beyond aggregate performance metrics and motivating more adaptive mechanisms.

major comments (1)

[Evaluation section / abstract] The central claim that 'controlled evaluations reveal clear coordination trade-offs' is load-bearing but unsupported in detail: no quantitative results, error bars, data-exclusion rules, or experimental protocol are supplied, and it is unclear whether a common agent substrate was used or whether planning algorithms, observation models, reward shaping, event generation rates, and constraint tightness were held fixed across paradigms (see § on evaluations and the abstract's description of the four paradigms). Without explicit controls, reported differences cannot be attributed to information flow, decision authority, and conflict resolution mechanisms alone.

minor comments (1)

[Abstract] The abstract asserts specific trade-offs without referencing any tables, figures, or quantitative values that demonstrate them.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their constructive feedback, which highlights important areas for strengthening the clarity and rigor of our evaluation methodology. We address the major comment point by point below and will revise the manuscript to incorporate the requested details.

read point-by-point responses

Referee: [Evaluation section / abstract] The central claim that 'controlled evaluations reveal clear coordination trade-offs' is load-bearing but unsupported in detail: no quantitative results, error bars, data-exclusion rules, or experimental protocol are supplied, and it is unclear whether a common agent substrate was used or whether planning algorithms, observation models, reward shaping, event generation rates, and constraint tightness were held fixed across paradigms (see § on evaluations and the abstract's description of the four paradigms). Without explicit controls, reported differences cannot be attributed to information flow, decision authority, and conflict resolution mechanisms alone.

Authors: We agree that the evaluation section would benefit from more explicit documentation to fully substantiate the controlled nature of the experiments. In the revised manuscript, we will expand the relevant section to include quantitative results (means and standard deviations across repeated trials) with error bars, explicit data-exclusion rules, and a detailed experimental protocol. All four paradigms were evaluated on a common agent substrate with planning algorithms, observation models, reward shaping, event generation rates, and constraint tightness held fixed; only the coordination mechanisms (information flow, decision authority, and conflict resolution) were varied. We will add a dedicated subsection and table that explicitly lists these fixed parameters alongside the varying coordination structures, enabling readers to attribute observed differences directly to the paradigms under study. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical benchmark comparison with independent evaluations

full rationale

The paper introduces DESBench as an empirical benchmark for comparing four coordination paradigms (centralized, hierarchical, heterarchical, holonic) in event-driven scheduling. The central claims consist of observed trade-offs in robustness, efficiency, misalignment, and constraint satisfaction drawn from controlled evaluations. No derivations, equations, fitted parameters, or predictions are present that reduce to quantities defined inside the paper. The work contains no self-citation load-bearing steps, uniqueness theorems, or ansatzes; all reported results are external to any internal definitions and rest on the benchmark tasks and metrics themselves. This is a standard empirical study whose findings are falsifiable against the described environment and do not exhibit any of the enumerated circularity patterns.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central claim rests on the domain assumption that the four paradigms represent distinct, representative mechanisms of coordination and that the simulated environment faithfully models industrial constraints; no free parameters or new physical entities are introduced.

axioms (1)

domain assumption The four coordination paradigms (centralized, hierarchical, heterarchical, holonic) correspond to distinct mechanisms of information flow, decision authority, and conflict resolution.
Invoked when defining the benchmark tasks and when interpreting the observed trade-offs.

invented entities (1)

DESBench no independent evidence
purpose: Shared discrete-event environment for evaluating agent coordination under partial observability and dynamically coupled constraints.
Newly introduced benchmark; no independent external validation or falsifiable prediction outside the paper is mentioned.

pith-pipeline@v0.9.0 · 5567 in / 1397 out tokens · 49112 ms · 2026-05-14T02:14:57.771418+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear
We focus on four representative coordination paradigms: centralized, hierarchical, heterarchical, and holonic... Our controlled evaluations reveal clear coordination trade-offs
IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear
effectiveness, constraint alignment, coordination efficiency, and robustness

Reference graph

Works this paper leans on

41 extracted references · 41 canonical work pages

[1]

Agentic services computing

Shuiguang Deng, Hailiang Zhao, Ziqi Wang, Guanjie Cheng, Peng Chen, Wenzhuo Qian, Zhiwei Ling, Jianwei Yin, Albert Y Zomaya, and Schahram Dustdar. Agentic services computing. arXiv preprint arXiv:2509.24380, 2025

work page arXiv 2025
[2]

A survey on llm-based multi-agent systems: workflow, infrastructure, and challenges.Vicinagearth, 1(1):9, 2024

Xinyi Li, Sai Wang, Siqi Zeng, Yu Wu, and Yi Yang. A survey on llm-based multi-agent systems: workflow, infrastructure, and challenges.Vicinagearth, 1(1):9, 2024

work page 2024
[3]

Overcoming the sim-to-real gap: Leveraging simulation to learn to explore for real-world rl

Andrew Wagenmaker, Kevin Huang, Liyiming Ke, Kevin Jamieson, and Abhishek Gupta. Overcoming the sim-to-real gap: Leveraging simulation to learn to explore for real-world rl. Advances in Neural Information Processing Systems, 37:78715–78765, 2024

work page 2024
[4]

Modeling complex system dynamics with flow matching across time and conditions

Martin Rohbeck, Edward De Brouwer, Charlotte Bunne, Jan-Christian Huetter, Anne Biton, Kelvin Y Chen, Aviv Regev, and Romain Lopez. Modeling complex system dynamics with flow matching across time and conditions. InThe Thirteenth International Conference on Learning Representations, 2025

work page 2025
[5]

Multiagentbench: Evaluating the collaboration and competition of llm agents

Kunlun Zhu, Hongyi Du, Zhaochen Hong, Xiaocheng Yang, Shuyi Guo, Daisy Zhe Wang, Zhenhailong Wang, Cheng Qian, Robert Tang, Heng Ji, et al. Multiagentbench: Evaluating the collaboration and competition of llm agents. InProceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 8580–8622, 2025

work page 2025
[6]

Battleagentbench: A benchmark for evaluating cooperation and competition capabilities of language models in multi-agent systems, 2024

Wei Wang, Dan Zhang, Tao Feng, Boyan Wang, and Jie Tang. Battleagentbench: A benchmark for evaluating cooperation and competition capabilities of language models in multi-agent systems, 2024. URLhttps://arxiv.org/abs/2408.15971

work page arXiv 2024
[7]

Multi-agent coordination across diverse applications: A survey.arXiv preprint arXiv:2502.14743, 2025

Lijun Sun, Yijun Yang, Qiqi Duan, Yuhui Shi, Chao Lyu, Yu-Cheng Chang, Chin-Teng Lin, and Yang Shen. Multi-agent coordination across diverse applications: A survey.arXiv preprint arXiv:2502.14743, 2025

work page arXiv 2025
[8]

Multi- agent collaboration via evolving orchestration.Advances in Neural Information Processing Systems, 2025

Yufan Dang, Chen Qian, Xueheng Luo, Jingru Fan, Zihao Xie, Ruijie Shi, Weize Chen, Cheng Yang, Xiaoyin Che, Ye Tian, Xuantang Xiong, Lei Han, Zhiyuan Liu, and Maosong Sun. Multi- agent collaboration via evolving orchestration.Advances in Neural Information Processing Systems, 2025

work page 2025
[9]

Multi-agent guided policy optimization.arXiv preprint arXiv:2507.18059, 2025

Yueheng Li, Guangming Xie, and Zongqing Lu. Multi-agent guided policy optimization.arXiv preprint arXiv:2507.18059, 2025

work page arXiv 2025
[10]

Central- ized permutation equivariant policy for cooperative multi-agent reinforcement learning.arXiv preprint arXiv:2508.11706, 2025

Zhuofan Xu, Benedikt Bollig, Matthias Függer, Thomas Nowak, and Vincent Le Dréau. Central- ized permutation equivariant policy for cooperative multi-agent reinforcement learning.arXiv preprint arXiv:2508.11706, 2025

work page arXiv 2025
[11]

Multi-agent coordination via multi-level communication.Advances in Neural Information Processing Systems, 37:118513–118539, 2024

Ziluo Ding, Zeyuan Liu, Zhirui Fang, Kefan Su, Liwen Zhu, and Zongqing Lu. Multi-agent coordination via multi-level communication.Advances in Neural Information Processing Systems, 37:118513–118539, 2024

work page 2024
[12]

Hm-rag: Hierarchical multi-agent multimodal retrieval augmented generation

Pei Liu, Xin Liu, Ruoyu Yao, Junming Liu, Siyuan Meng, Ding Wang, and Jun Ma. Hm-rag: Hierarchical multi-agent multimodal retrieval augmented generation. InProceedings of the 33rd ACM international conference on multimedia, pages 2781–2790, 2025

work page 2025
[13]

Hiva: Self-organized hierarchical variable agent via goal-driven semantic-topological evolution

Jinzhou Tang, Jusheng Zhang, Qinhan Lv, Sidi Liu, Jing Yang, Chengpei Tang, and Keze Wang. Hiva: Self-organized hierarchical variable agent via goal-driven semantic-topological evolution. InProceedings of the AAAI Conference on Artificial Intelligence, volume 40, pages 2083–2091, 2026

work page 2083
[14]

Hierarchical decision making based on structural information principles.Journal of Machine Learning Research, 26(182):1–55, 2025

Xianghua Zeng, Hao Peng, Dingli Su, and Angsheng Li. Hierarchical decision making based on structural information principles.Journal of Machine Learning Research, 26(182):1–55, 2025

work page 2025
[15]

Agentnet: Decentralized evolutionary coordination for llm-based multi-agent systems

Yingxuan Yang, Huacan Chai, Shuai Shao, Yuanyi Song, Siyuan Qi, Renting Rui, and Weinan Zhang. Agentnet: Decentralized evolutionary coordination for llm-based multi-agent systems. Advances in Neural Information Processing Systems, 2025. 10

work page 2025
[16]

Comas: Co-evolving multi-agent systems via interaction rewards

Xiangyuan Xue, Yifan Zhou, Guibin Zhang, Zaibin Zhang, Yijiang Li, Chen Zhang, Zhenfei Yin, Philip Torr, Wanli Ouyang, and Lei Bai. Comas: Co-evolving multi-agent systems via interaction rewards. InThe Fourteenth International Conference on Learning Representations, 2026

work page 2026
[17]

Hygma: Hypergraph coordination networks with dynamic grouping for multi-agent reinforcement learning.arXiv preprint arXiv:2505.07207, 2025

Chiqiang Liu and Dazi Li. Hygma: Hypergraph coordination networks with dynamic grouping for multi-agent reinforcement learning.arXiv preprint arXiv:2505.07207, 2025

work page arXiv 2025
[18]

Goutham Nalagatla. Hierarchical decentralized multi-agent coordination with privacy- preserving knowledge sharing: Extending agentnet for scalable autonomous systems.arXiv preprint arXiv:2512.00614, 2025

work page arXiv 2025
[19]

Contextual knowledge sharing in multi-agent reinforcement learning with decentralized communication and coordination.arXiv preprint arXiv:2501.15695, 2025

Hung Du, Srikanth Thudumu, Hy Nguyen, Rajesh Vasa, and Kon Mouzakis. Contextual knowledge sharing in multi-agent reinforcement learning with decentralized communication and coordination.arXiv preprint arXiv:2501.15695, 2025

work page arXiv 2025
[20]

Hmlb: Holonic multi-agent approach for preventive controllers load-balancing in sdn-enabled smart grid.Computer Communications, 228:107984, 2024

Marjan Keramati, Sauleh Etemedi, and Nasser Mozayani. Hmlb: Holonic multi-agent approach for preventive controllers load-balancing in sdn-enabled smart grid.Computer Communications, 228:107984, 2024

work page 2024
[21]

Agentsnet: Coordination and collaborative reasoning in multi-agent llms.arXiv preprint arXiv:2507.08616, 2025

Florian Grötschla, Luis Müller, Jan Tönshoff, Mikhail Galkin, and Bryan Perozzi. Agentsnet: Coordination and collaborative reasoning in multi-agent llms.arXiv preprint arXiv:2507.08616, 2025

work page arXiv 2025
[22]

Gem- mas: Graph-based evaluation metrics for multi agent systems

Jisoo Lee, Raeyoung Chang, Dongwook Kwon, Harmanpreet Singh, and Nikhil Verma. Gem- mas: Graph-based evaluation metrics for multi agent systems. InProceedings of the 2025 Conference on Empirical Methods in Natural Language Processing: Industry Track, pages 1522–1532, 2025

work page 2025
[23]

Llm-based multi-agent systems: Frameworks, evaluation, open chal- lenges, and research frontiers

Soharab Hossain Shaikh. Llm-based multi-agent systems: Frameworks, evaluation, open chal- lenges, and research frontiers. InInternational Joint Conference on Computational Intelligence, pages 149–170. Springer, 2025

work page 2025
[24]

Bono, Brendan D

Kunal Menda, Yi-Chun Chen, Justin Grana, James W. Bono, Brendan D. Tracey, Mykel J. Kochenderfer, and David Wolpert. Deep reinforcement learning for event-driven multi-agent decision processes.IEEE Transactions on Intelligent Transportation Systems, 20(4):1259–1268,

work page
[25]

doi: 10.1109/TITS.2018.2848264

work page doi:10.1109/tits.2018.2848264 2018
[26]

Moseac: Streamlined variable time step reinforcement learning.arXiv preprint arXiv:2406.01521, 2024

Dong Wang and Giovanni Beltrame. Moseac: Streamlined variable time step reinforcement learning.arXiv preprint arXiv:2406.01521, 2024

work page arXiv 2024
[27]

Benchmarking flexible job-shop scheduling and control systems.Control Engineering Practice, 21(9):1204–1225, 2013

Damien Trentesaux, Cyrille Pach, Abdelghani Bekrar, Yves Sallez, Thierry Berger, Thérèse Bonte, Paulo Leitão, and José Barbosa. Benchmarking flexible job-shop scheduling and control systems.Control Engineering Practice, 21(9):1204–1225, 2013. ISSN 0967-0661. doi: https://doi.org/10.1016/j.conengprac.2013.05.004. URL https://www.sciencedirect. com/science/...

work page doi:10.1016/j.conengprac.2013.05.004 2013
[28]

Proposition of an implementation framework enabling benchmarking of holonic manufacturing systems.CoRR, abs/1901.05669, 2019

Olivier Cardin and Anne L’Anton. Proposition of an implementation framework enabling benchmarking of holonic manufacturing systems.CoRR, abs/1901.05669, 2019. URL http: //arxiv.org/abs/1901.05669

work page arXiv 1901
[29]

A production scheduling framework for reinforcement learning under real-world constraints

Jonathan Hoss, Felix Schelling, and Noah Klarmann. A production scheduling framework for reinforcement learning under real-world constraints. In2025 IEEE 21st International Conference on Automation Science and Engineering (CASE), pages 1736–1743. IEEE, 2025

work page 2025
[30]

Agent ai with langgraph: A modular framework for enhancing machine translation using large language models.arXiv preprint arXiv:2412.03801, 2024

Jialin Wang and Zhihua Duan. Agent ai with langgraph: A modular framework for enhancing machine translation using large language models.arXiv preprint arXiv:2412.03801, 2024

work page arXiv 2024
[31]

Agentscope: A flexible yet robust multi-agent platform.arXiv preprint arXiv:2402.14034, 2024

Dawei Gao, Zitao Li, Xuchen Pan, Weirui Kuang, Zhijian Ma, Bingchen Qian, Fei Wei, Wenhao Zhang, Yuexiang Xie, Daoyuan Chen, et al. Agentscope: A flexible yet robust multi-agent platform.arXiv preprint arXiv:2402.14034, 2024

work page arXiv 2024
[32]

A taxonomy of hierarchical multi-agent systems: Design patterns, coordination mechanisms, and industrial applications.arXiv preprint arXiv:2508.12683, 2025

David J Moore. A taxonomy of hierarchical multi-agent systems: Design patterns, coordination mechanisms, and industrial applications.arXiv preprint arXiv:2508.12683, 2025. 11

work page arXiv 2025
[33]

Sequential cooperative multi-agent online learning and adaptive coordination control in dynamic and uncertain environments

Limengxi Yue, Duo Xu, Dong Qiu, Yanpei Shi, Shuyang Xu, and Manish Shah. Sequential cooperative multi-agent online learning and adaptive coordination control in dynamic and uncertain environments. In2025 5th International Conference on Electronic Information Engineering and Computer Communication (EIECC), pages 692–697. IEEE, 2025

work page 2025
[34]

Cp-agentnet: Autonomous and explainable communication protocol design using generative agents

Dae Cheol Kwon and Xinyu Zhang. Cp-agentnet: Autonomous and explainable communication protocol design using generative agents. In2025 IEEE 33rd International Conference on Network Protocols (ICNP), pages 1–12. IEEE, 2025

work page 2025
[35]

Amas: Adaptively determining communication topology for llm-based multi-agent system

Hui Yi Leong, Yuheng Li, Yuqing Wu, Wenwen Ouyang, Wei Zhu, Jiechao Gao, and Wei Han. Amas: Adaptively determining communication topology for llm-based multi-agent system. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing: Industry Track, pages 2061–2070, 2025. A Additional Environment and Runtime Details One decisi...

work page 2025
[36]

Starting from the post-decision state x+ τk, the event engine advances autonomously until the next decision-relevant stopping point is reached, yielding a new pre-decision state x− τk+1 together with realized world eventsE k+1

work page
[37]

The event interpreter combines these world events with any newly materialized protocol-level objectsM k+1 and computes the next activated setU(τ k+1)

work page
[38]

For each activated agenti∈ U(τ k+1), the runtime constructs the benchmark-visible decision boundary (local observation), a protocol context, and a protocol-conditioned legal action set

work page
[39]

A.1 Event Engine The event engine is the shared discrete-event scheduling substrate used by all coordination protocols

Returned actions are validated, committed to the event engine, and written back as updated state and protocol artifacts. A.1 Event Engine The event engine is the shared discrete-event scheduling substrate used by all coordination protocols. It advances physical production time, realizes environment events, updates job, machine, buffer, transport, blocking...

work page
[40]

World-event activations, induced by newly realized environment events such as release op- portunities, completion-triggered routing, queue-state changes, budget alarms, breakdowns, or repairs

work page
[41]

agent":

Protocol-object activations, induced by coordination objects addressed to a specific agent, such as assignment requests, parent-to-child allocations, child feedback, settlement notices, rejection records, or escalation requests. Only agents in U(τ k) are required to act at epoch τk, and the activated set itself is an output of the event interpreter. Contr...

work page 1966