betaFeedback

WorksComputer Science

Learning Multi-Agent Orchestration:Higher-Order Interactions in Policy Improvement

Abstract

Learned orchestration has emerged as a promising paradigm for large language model (LLM)-based multi-agent systems, in which a central orchestrator selects agents and controls the information they receive throughout a task. Despite rapid empirical progress, existing methods train the orchestrator by crediting each decision with its own advantage, implicitly assuming that the other decisions in the episode stay fixed. This assumption fails in orchestration, where the contribution of each decision to the final outcome depends on the decisions preceding it, so the gain of a joint update is not the sum of its per-decision effects. To address this challenge, we introduce IAPO, an interaction-aware policy optimization framework that explicitly accounts for interactions among decisions. Specifically, IAPO exactly decomposes the performance difference between two policies into the effects of individual decisions and their interactions, which we prove can account for half of the attainable gain, and then optimizes both from sampled trajectories while certifying that every update is an improvement. On long-input versions of GSM8K, MATH, DROP, MBPP and HumanEval, against fifteen published multi-agent and workflow-search systems on the same 7B executor, IAPO's deployed plans reach the strongest system's accuracy with fewer tokens, and removing the interaction term costs 7 to 17 points of accuracy on code.

Introduction

Large language model (LLM)-based agents have achieved strong results in multi-step reasoning [44], code generation [33], and tool use [28]. Organizing several such agents into a multi-agent system extends these capabilities further, as agents that debate [8], role-play [20], or converse [42] often outperform each of their members on complex tasks. This benefit, however, depends on how the collaboration is organized: which agent acts next, which artifacts it may read, and when the process terminates [46]. What an agent reads also determines how much it influences the others, and hence what a population of interacting agents comes to believe collectively [21].

Early multi-agent systems specify this organization manually through fixed role assignments and communication pipelines [12, 42], which limits their transfer across tasks and domains. Subsequent work automates parts of the design: GPTSwarm [52] and G-Designer [47] optimize communication topologies, MasRouter [45] learns to route queries among agents, and ADAS [13], AFlow [49], and MAS [48] search over entire workflows. Nevertheless, these methods largely fix the organization before execution and cannot adapt it to the intermediate results produced during a task. More recently, orchestration has been cast as a sequential decision problem, in which a central orchestrator observes the evolving state of a task and is trained by reinforcement learning (RL) to decide how to proceed at every step [7, 25]. The same principle has been extended to training the coordinated agents jointly with the orchestrator [22, 50], producing systems that surpass the strongest individual agents at reduced inference cost [7, 25].

Although this paradigm has shown clear empirical gains, its training objectives are inherited directly from single-agent policy optimization [30, 31], which credits each decision with its own advantage while treating all other decisions in the episode as fixed. This approximation is justified by trust-region arguments that bound its error through the per-step policy divergence [15, 29], and multi-agent extensions control the error in the same per-step manner [18, 51]. Orchestration violates the premise of these arguments in two ways. First, the decisions within an episode depend on one another, since the effect of each decision on the outcome changes with the decisions made before it. A joint change to several decisions can therefore improve performance even when every single change degrades it, and an objective that adds up per-decision credits cannot detect such joint improvements. Second, an orchestrator makes many decisions in one episode. Even if each decision changes only slightly, the errors add up across decisions, so a per-step trust region no longer keeps the approximation accurate over the length of a whole episode.

To address these challenges, we develop an interaction theory of policy improvement for orchestration and propose IAPO (Interaction-Aware Policy Optimization). We show that the performance difference between two policies is a finite sum of interaction terms among decisions. Each term can be estimated from trajectories without a critic, and the existing objective is the first-order term. This objective has the correct gradient but misses all interactions between organizational decisions, such as which worker to call and which outputs it may read, so any update rule that treats these decisions separately can lose half of the possible gain. We further bound each interaction term by how often the decisions involved are made, which implies that a trust region should limit the policy change weighted by visit counts. Based on this theory, IAPO optimizes the first- and second-order terms in one pass over each trajectory and accepts an update only if a high-confidence lower bound on its gain is positive. We prove that every accepted update improves performance with high probability. On controlled orchestration tasks, IAPO reaches the best possible gain on all three tasks, while the existing objective fails on two, and IAPO never accepts a harmful update. On long-input versions of five benchmarks, against fifteen published multi-agent and workflow-search systems on the same 7B executor, IAPO’s deployed plans reach the strongest system’s accuracy with fewer tokens in every environment and, at that system’s token end_mid=1 budget, exceed it significantly on GSM8K, MATH and DROP under at least one cost weight; removing the interaction term costs 7 to 17 points of accuracy on code.

  • Exact Interaction Theory: We derive an exact, finite expansion of the performance difference between two orchestration policies, which identifies existing objectives as its single-decision terms and characterizes the omitted interactions through their curvature, visit-count bounds, and sample complexity.

  • Practical Algorithm: We propose IAPO, which jointly optimizes the single-decision and pairwise terms in one pass per trajectory and accepts an update only when its gain is certified with high confidence. Its analysis extends to jointly trained agents.

  • Theoretical and Empirical Validation: We prove that ignoring interactions can forfeit half of the attainable gain and that IAPO accepts only improving updates with high probability, and we confirm on controlled tasks that IAPO recovers gains missed by existing objectives without accepting harmful updates, and on five LLM orchestration benchmarks that it matches or exceeds the strongest of fifteen baselines at lower cost.

Problem Formulation

In this section, we formulate multi-agent orchestration as an episodic MDP and define our optimization objective.

Multi-Agent Orchestration

Let Q\mathcal{Q} be a finite set of tasks and Y\mathcal{Y} a finite set of outputs, such as text responses. A task q∈Qq \in\mathcal{Q} is drawn from a distribution D\mathcal{D} and solved by the agents N={0,1,…,M}\mathcal{N} = \{0, 1, \ldots, M\}, where agent 0 is the orchestrator and agents j∈[M]j \in[M] are workers, such as LLMs, tools, and verifiers. The orchestrator solves the task through a sequence of worker calls. Let y1,…,yk∈Yy_1, \ldots, y_k \in\mathcal{Y} be the outputs of the first kk calls. In the (k+1)(k+1)-th call, the orchestrator chooses a worker j∈[M]j \in[M] and an input set E⊆{y1,…,yk}\mathcal{E} \subseteq\{y_1, \ldots, y_k\} of earlier outputs, and worker jj reads task qq and E\mathcal{E} and returns a new output yk+1∈Yy_{k+1} \in\mathcal{Y}. Each call to worker jj costs cj≥0c_j \ge0, and at most KK calls are allowed. Instead of calling a worker, the orchestrator may stop, and the task and the outputs collected so far are scored by a task reward R∈[0,1]R \in[0,1].

Orchestration MDP

We model multi-agent orchestration as an episodic MDP M(S,A,H,P,r,d0)\mathcal{M}(S, A, H, P, r, d_0), where SS is the state space, AA is the action space, HH is the number of steps in each episode, P={Pt}t=1HP = \{P_t\}_{t=1}^{H} with Pt(⋅∣s,a)P_t(\cdot\mid s, a) the distribution of the next state when action aa is taken at state ss in step tt, r={rt}t=1Hr = \{r_t\}_{t=1}^{H} with rt:S×A→Rr_t : S \times A \to\mathbb{R} the reward function at step tt, and d0=Dd_0 = \mathcal{D} the initial state distribution.

Episode dynamics. An episode starts from the task, s1=q∼Ds_1 = q \sim\mathcal{D}, and each agent i∈Ni \in\mathcal{N} selects actions with its own policy πi\pi^i. After kk calls, the interaction history hk=((j1,E1,y1),…,(jk,Ek,yk))h_k = ((j_1, \mathcal{E}_1, y_1), \ldots, (j_k, \mathcal{E}_k, y_k)) records each call as a triple: in the ℓ\ell-th call, worker jℓ∈[M]j_\ell\in[M] receives the set Eℓ⊆Yℓ−1\mathcal{E}_\ell\subseteq\mathcal{Y}_{\ell-1} of earlier outputs and returns yℓ∈Yy_\ell\in\mathcal{Y}, where Yk={y1,…,yk}\mathcal{Y}_k = \{y_1, \ldots, y_k\}. Each call takes two steps.

  1. Orchestrator step. At state st=(q,hk)s_t = (q, h_k), the orchestrator acts, it=0i_t = 0. It decides whether to continue, and if so, which worker j∈[M]j \in[M] to call and which outputs E⊆Yk\mathcal{E} \subseteq\mathcal{Y}_k of earlier workers to give it. Formally, it samples at∼π0(⋅∣st)a_t \sim\pi^0(\cdot\mid s_t) from:

A(st)={(j,E):j∈[M],E⊆Yk}∪{stop}.\mathcal{A}(s_t) = \{(j,\mathcal{E}) : j \in[M], \mathcal{E} \subseteq\mathcal{Y}_k\} \cup\{\mathrm{stop}\}.

A call (j,E)(j,\mathcal{E}) receives reward rt=−λcjr_t = -\lambda c_j with cost coefficient λ≥0\lambda\ge0 and moves the state to st+1=(q,hk,j,E)s_{t+1} = (q, h_k, j, \mathcal{E}), and stop receives the task reward rt=R(q,hk)r_t = R(q, h_k) and ends the episode.

  1. Worker step. Worker jj then receives the call from the orchestrator at state st+1=(q,hk,j,E)s_{t+1} = (q, h_k, j, \mathcal{E}), so it+1=ji_{t+1} = j. It reads the task qq and the outputs in E\mathcal{E} and produces a new output. Formally, it samples at+1=y∼πj(⋅∣q,E)a_{t+1} = y \sim\pi^j(\cdot\mid q,\mathcal{E}) from Y\mathcal{Y} and receives rt+1=0r_{t+1} = 0. The output is added to the history, hk+1=(hk,(j,E,y))h_{k+1} = (h_k, (j,\mathcal{E},y)), and control returns to the orchestrator at state st+2=(q,hk+1)s_{t+2} = (q, h_{k+1}).

The joint policy is therefore πt(a∣st)=πit(a∣st)\pi_t(a \mid s_t) = \pi^{i_t}(a \mid s_t), and every trajectory (s1,a1,…,sH,aH)(s_1, a_1, \ldots, s_H, a_H) is generated by π\pi.

Policies and values. A policy π={πt}t=1H\pi= \{\pi_t\}_{t=1}^{H} maps each state ss to a distribution πt(⋅∣s)\pi_t(\cdot\mid s) over A(s)\mathcal{A}(s). The value and action-value functions of π\pi are:

Vtπ(s)=Eπ[∑t′=tHrt′(st′,at′) | st=s],V_t^\pi(s) = \mathbb{E}_\pi\left[\sum_{t'=t}^{H} r_{t'}(s_{t'}, a_{t'}) \,\middle|\, s_t = s\right],
Qtπ(s,a)=Eπ[∑t′=tHrt′(st′,at′) | st=s,at=a],Q_t^\pi(s,a) = \mathbb{E}_\pi\left[\sum_{t'=t}^{H} r_{t'}(s_{t'}, a_{t'}) \,\middle|\, s_t = s, a_t = a\right],

and the advantage function is Atπ=Qtπ−VtπA_t^\pi= Q_t^\pi- V_t^\pi. The return-to-go of a trajectory from step tt is Gt=∑t′=tHrt′(st′,at′)G_t = \sum_{t'=t}^{H} r_{t'}(s_{t'}, a_{t'}).

Learning Objective

Our goal is to solve each task with high reward at low cost. We keep the worker policies π1,…,πM\pi^1,\ldots,\pi^M fixed and optimize the orchestrator policy π0\pi^0 to maximize the expected task reward minus the cost of calls:

max⁡π0J(π):=Eπ[R(q,h)−λ∑(j,E,y)∈hcj].(1)\max_{\pi^0} J(\pi) := \mathbb{E}_\pi\left[R(q,h) - \lambda\sum_{(j,E,y)\in h} c_j\right]. \tag*{(1)}

References

  1. [1]Jacob Austin, Augustus Odena, Maxwell Nye, Maarten Bosma, Henryk Michalewski, David Dohan, Ellen Jiang, Carrie Cai, Michael Terry, Quoc Le, and Charles Sutton. Program synthesis with large language models. arXiv preprint arXiv:2108.07732, 2021.arxiv.org/abs/2108.07732
  2. [2]Federico Cassano, John Gouwar, Daniel Nguyen, Sydney Nguyen, Luna Phipps-Costin, Donald Pinckney, Ming-Ho Yee, Yangtian Zi, Carolyn Jane Anderson, Molly Q. Feldman, Arjun Guha, Michael Greenberg, and Abhinav Jangda. MultiPL-E: A scalable and polyglot approach to benchmarking neural code generation. IEEE Transactions on Software Engineering, 49(7), 2023.DOI
  3. [3]Guangyao Chen, Siwei Dong, Yu Shu, Ge Zhang, Jaward Sesay, Börje F. Karlsson, Jie Fu, and Yemin Shi. AutoAgents: A framework for automatic agent generation. In Proceedings of the Thirty-Third International Joint Conference on Artificial Intelligence, 2024.arxiv.org/abs/2309.17288
  4. [4]Mark Chen, Jerry Tworek, Heewoo Jun, Qiming Yuan, et al. Evaluating large language models trained on code. arXiv preprint arXiv:2107.03374, 2021.arxiv.org/abs/2107.03374
  5. [5]Weize Chen, Yusheng Su, Jingwei Zuo, Cheng Yang, Chenfei Yuan, Chi-Min Chan, Yujia Qin, Yaxi Lu, Ruobing Xie, et al. AgentVerse: Facilitating multi-agent collaboration and exploring emergent behaviors in agents. arXiv preprint arXiv:2308.10848, 2023.arxiv.org/abs/2308.10848
  6. [6]Karl Cobbe, Vineet Kosaraju, Mohammad Bavarian, Mark Chen, Heewoo Jun, Qiming Yuan, et al. Training verifiers to solve math word problems. arXiv preprint arXiv:2110.14168, 2021.arxiv.org/abs/2110.14168
  7. [7]Yufan Dang, Chen Qian, Xueheng Luo, Jingru Fan, Zihao Xie, Ruijie Shi, Weize Chen, Cheng Yang, Xiaoyin Che, Ye Tian, et al. Multi-agent collaboration via evolving orchestration. Advances in neural information processing systems, 38:165025–165059, 2026.arxiv.org/abs/2505.19591
  8. [8]Yilun Du, Shuang Li, Antonio Torralba, Joshua B Tenenbaum, and Igor Mordatch. Improving factuality and reasoning in language models through multiagent debate. arXiv preprint arXiv:2305.14325, 2023.arxiv.org/abs/2305.14325
  9. [9]Dheeru Dua, Yizhong Wang, Pradeep Dasigi, Gabriel Stanovsky, Sameer Singh, and Matt Gardner. DROP: A reading comprehension benchmark requiring discrete reasoning over paragraphs. In Proceedings of NAACL-HLT, 2019.arxiv.org/abs/1903.00161
  10. [10]Yao Fu, Hao Peng, Ashish Sabharwal, Peter Clark, and Tushar Khot. Complexity-based prompting for multi-step reasoning. In International Conference on Learning Representations, 2023.arxiv.org/abs/2210.00720
  11. [11]Dan Hendrycks, Collin Burns, Saurav Kadavath, Akul Arora, Steven Basart, Eric Tang, Dawn Song, and Jacob Steinhardt. Measuring mathematical problem solving with the MATH dataset. In Advances in Neural Information Processing Systems, Datasets and Benchmarks Track, 2021.
  12. [12]Sirui Hong, Mingchen Zhuge, Jonathan Chen, Xiawu Zheng, Yuheng Cheng, Jinlin Wang, Ceyao Zhang, Steven Yau, Zijuan Lin, Liyang Zhou, et al. MetaGPT: Meta programming for a multi-agent collaborative framework. In International Conference on Learning Representations, volume 2024, pages 23247–23275, 2024.arxiv.org/abs/2308.00352
  13. [13]Shengran Hu, Cong Lu, and Jeff Clune. Automated design of agentic systems. In International Conference on Learning Representations, volume 2025, pages 21344–21377, 2025.arxiv.org/abs/2408.08435
  14. [14]Dongfu Jiang, Xiang Ren, and Bill Yuchen Lin. LLM-Blender: Ensembling large language models with pairwise ranking and generative fusion. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics, 2023.arxiv.org/abs/2306.02561
  15. [15]Sham Kakade and John Langford. Approximately optimal approximate reinforcement learning. In Proceedings of the nineteenth international conference on machine learning, pages 267–274, 2002.
  16. [16]Zohar Karnin, Tomer Koren, and Oren Somekh. Almost optimal exploration in multi-armed bandits. In International Conference on Machine Learning, pages 1238–1246. PMLR, 2013.
  17. [17]Zixuan Ke, Austin Xu, Yifei Ming, Xuan-Phi Nguyen, Caiming Xiong, and Shafiq Joty. MAS-Zero: Designing multi-agent systems with zero supervision, 2025. Preprint.arxiv.org/abs/2505.14996
  18. [18]Jakub Grudzien Kuba, Ruijing Chen, Muning Wen, Ying Wen, Fanglei Sun, Jun Wang, and Yaodong Yang. Trust region policy optimisation in multi-agent reinforcement learning. arXiv preprint arXiv:2109.11251, 2021.arxiv.org/abs/2109.11251
  19. [19]Woosuk Kwon, Zhuohan Li, Siyuan Zhuang, Ying Sheng, Lianmin Zheng, Cody Hao Yu, Joseph Gonzalez, Hao Zhang, and Ion Stoica. Efficient memory management for large language model serving with PagedAttention. In Proceedings of the 29th Symposium on Operating Systems Principles, 2023.arxiv.org/abs/2309.06180
  20. [20]Guohao Li, Hasan Hammoud, Hani Itani, Dmitrii Khizbullin, and Bernard Ghanem. Camel: Communicative agents for "mind" exploration of large language model society. Advances in neural information processing systems, 36: 51991–52008, 2023.arxiv.org/abs/2303.17760
  21. [21]Kaixuan Liu, Guojun Xiong, Weinan Zhang, and Shengpu Tang. Social networks of LLM agents. arXiv preprint arXiv:2607.03695, 2026a.arxiv.org/abs/2607.03695
  22. [22]Shuo Liu, Zeyu Liang, Xueguang Lyu, and Christopher Amato. Llm collaboration with multi-agent reinforcement learning. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 40, pages 32150–32158, 2026b.
  23. [23]Zijun Liu, Yanzhe Zhang, Peng Li, Yang Liu, and Diyi Yang. Dynamic LLM-agent network: An LLM-agent collaboration framework with agent team optimization, 2023. Preprint.
  24. [24]Andreas Maurer and Massimiliano Pontil. Empirical bernstein bounds and sample variance penalization. arXiv preprint arXiv:0907.3740, 2009.arxiv.org/abs/0907.3740
  25. [25]Stefan Nielsen, Edoardo Cetin, Peter Schwendeman, Qi Sun, Jinglue Xu, and Yujin Tang. Learning to orchestrate agents in natural language with the conductor. In International Conference on Learning Representations, volume 2026, pages 135686–135724, 2026.arxiv.org/abs/2512.04388
  26. [26]Chen Qian, Zihao Xie, YiFei Wang, Wei Liu, Kunlun Zhu, Hanchen Xia, Yufan Dang, Zhuoyun Du, Weize Chen, Cheng Yang, Zhiyuan Liu, and Maosong Sun. Scaling large language model-based multi-agent collaboration. In International Conference on Learning Representations, 2025.arxiv.org/abs/2406.07155
  27. [27]Stephen Robertson and Hugo Zaragoza. The probabilistic relevance framework: BM25 and beyond. Foundations and Trends in Information Retrieval, 3(4), 2009.
  28. [28]Timo Schick, Jane Dwivedi-Yu, Roberto Dessi, Roberta Riaileanu, Maria Lomeli, Eric Hambro, Luke Zettlemoyer, Nicola Cancedda, and Thomas Scialom. Toolformer: Language models can teach themselves to use tools. Advances in neural information processing systems, 36:68539–68551, 2023.
  29. [29]John Schulman, Sergey Levine, Pieter Abbeel, Michael Jordan, and Philipp Moritz. Trust region policy optimization. In International conference on machine learning, pages 1889–1897. Pmlr, 2015.arxiv.org/abs/1502.05477
  30. [30]John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov. Proximal policy optimization algorithms. arXiv preprint arXiv:1707.06347, 2017.arxiv.org/abs/1707.06347
  31. [31]Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, YK Li, Yang Wu, et al. Deepseekmath: Pushing the limits of mathematical reasoning in open language models. arXiv preprint arXiv:2402.03300, 2024.
  32. [32]Freda Shi, Xinyun Chen, Kanishka Misra, Nathan Scales, David Dohan, Ed H. Chi, Nathanael Schärli, and Denny Zhou. Large language models can be easily distracted by irrelevant context. In Proceedings of the 40th International Conference on Machine Learning, 2023.
  33. [33]Noah Shinn, Federico Cassano, Ashwin Gopinath, Karthik Narasimhan, and Shunyu Yao. Reflexion: Language agents with verbal reinforcement learning. Advances in neural information processing systems, 36:8634–8652, 2023.
  34. [34]Yunhao Tang, Michal Valko, and Rémi Munos. Taylor expansion policy optimization. In International Conference on Machine Learning, pages 9397–9406. PMLR, 2020.arxiv.org/abs/2003.06259
  35. [35]Philip Thomas, Georgios Theocharous, and Mohammad Ghavamzadeh. High confidence policy improvement. In International Conference on Machine Learning, pages 2380–2388. PMLR, 2015.
  36. [36]Alexandre B Tsybakov and Vladimir Zaiats. Introduction to nonparametric estimation, volume 11. Springer, 2009.
  37. [37]Jean Ville. Étude critique de la notion de collectif. Gauthier-Villars, Paris, 1939.
  38. [38]Xuezhi Wang, Jason Wei, Dale Schuurmans, Quoc Le, Ed Chi, Sharan Narang, Aakanksha Chowdhery, and Denny Zhou. Self-consistency improves chain of thought reasoning in language models. In International Conference on Learning Representations, 2023.
  39. [39]Zhenhailong Wang, Shaoguang Mao, Wenshan Wu, Tao Ge, Furu Wei, and Heng Ji. Unleashing the emergent cognitive synergy in large language models: A task-solving agent through multi-persona self-collaboration. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics, 2024.
  40. [40]Ian Waudby-Smith and Aaditya Ramdas. Estimating means of bounded random variables by betting. Journal of the Royal Statistical Society Series B: Statistical Methodology, 86(1):1–27, 2024.
  41. [41]Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Brian Ichter, Fei Xia, Ed Chi, Quoc Le, and Denny Zhou. Chain-of-thought prompting elicits reasoning in large language models. In Advances in Neural Information Processing Systems, 2022.
  42. [42]Qingyun Wu, Gagan Bansal, Jieyu Zhang, Yiran Wu, Beibin Li, Erkang Zhu, Li Jiang, Xiaoyun Zhang, Shaokun Zhang, Jiale Liu, et al. Autogen: Enabling next-gen llm applications via multi-agent conversation. arXiv preprint arXiv:2308.08155, 2023.arxiv.org/abs/2308.08155
  43. [43]An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Haoran Wei, et al. Qwen2.5 technical report. arXiv preprint arXiv:2412.15115, 2024.arxiv.org/abs/2412.15115
  44. [44]Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik Narasimhan, and Yuan Cao. React: Synergizing reasoning and acting in language models. arXiv preprint arXiv:2210.03629, 2022.arxiv.org/abs/2210.03629
  45. [45]Yanwei Yue, Guibin Zhang, Boyang Liu, Guancheng Wan, Kun Wang, Dawei Cheng, and Yiyan Qi. Masrouter: Learning to route llms for multi-agent systems. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 15549–15572, 2025.
  46. [46]Chenchen Zhang. Reinforcement learning for llm-based multi-agent systems through orchestration traces. arXiv preprint arXiv:2605.02801, 2026.arxiv.org/abs/2605.02801
  47. [47]Guibin Zhang, Yanwei Yue, Xiangguo Sun, Guancheng Wan, Miao Yu, Junfeng Fang, Kun Wang, Tianlong Chen, and Dawei Cheng. G-designer: Architecting multi-agent communication topologies via graph neural networks. arXiv preprint arXiv:2410.11782, 2024.arxiv.org/abs/2410.11782
  48. [48]Guibin Zhang, Luyang Niu, Junfeng Fang, Kun Wang, Lei Bai, and Xiang Wang. Multi-agent architecture search via agentic supernet. arXiv preprint arXiv:2502.04180, 2025a.arxiv.org/abs/2502.04180
  49. [49]Jiayi Zhang, Jinyu Xiang, Zhaoyang Yu, Fengwei Teng, Xionghui Chen, Jiaqi Chen, Mingchen Zhuge, Xin Cheng, Sirui Hong, Jinlin Wang, et al. Aflow: Automating agentic workflow generation. In International Conference on Learning Representations, volume 2025, pages 34040–34077, 2025b.arxiv.org/abs/2410.10762
  50. [50]Yujie Zhao, Lanxiang Hu, Yang Wang, Minmin Hou, Hao Zhang, Ke Ding, and Jishen Zhao. Stronger-mas: Multi-agent reinforcement learning for collaborative llms. In International Conference on Learning Representations, volume 2026, pages 150619–150651, 2026.arxiv.org/abs/2510.11062
  51. [51]Yifan Zhong, Jakub Grudzien Kuba, Xidong Feng, Siyi Hu, Jiaming Ji, and Yaodong Yang. Heterogeneous-agent reinforcement learning. Journal of Machine Learning Research, 25(32):1–67, 2024.arxiv.org/abs/2304.09870
  52. [52]Mingchen Zhuge, Wenyi Wang, Louis Kirsch, Francesco Faccio, Dmitrii Khizbullin, and Jürgen Schmidhuber. Language agents as optimizable graphs. arXiv preprint arXiv:2402.16823, 2024.arxiv.org/abs/2402.16823

Paper details

vlnc:W_b4dm5cwu
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 17

downloads · 1

A view counts once per browser per day, a download each time the PDF is requested.

  • Computer Science

Not recorded for this paper

View statement
Looking for related papers…
Contents