pith. sign in

arxiv: 2606.17368 · v1 · pith:ZKA2DJUCnew · submitted 2026-06-15 · 💻 cs.AI · cs.NI

Distributed General-Purpose Agent Networks: Architecture, Key Mechanisms, and Prototypes

Shengli Zhang , Deen Ma , Zibin Lin , Taotao Wang This is my paper

Pith reviewed 2026-06-27 02:55 UTC · model grok-4.3

classification 💻 cs.AI cs.NI
keywords distributed agent networkspeer-to-peer systemsmulti-agent systemsprotocol adaptationsemantic propagationreputation systemsmechanism designautonomous agents
0
0 comments X

The pith

Distributed agent networks require a dedicated protocol adaptation layer to propagate semantic declarations about intentions and capabilities.

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

The paper claims that open peer-to-peer networks of heterogeneous agents cannot be assembled simply by layering existing P2P overlays onto conventional multi-agent systems. A layered architecture built around a protocol adaptation layer is needed to move semantic information on intentions, capabilities, states, and cooperation constraints between task-level logic and network operations. This setup addresses three mechanism problems: propagating announcements for discovery, binding identities with reputation for governance, and generating mechanisms for task execution. Readers would care because individual agents remain limited by local data and permissions, while such networks could enable discovery and joint work across personal devices and edge nodes.

Core claim

The paper claims that distributed general-purpose agent networks, in which heterogeneous agents on personal devices, edge nodes, or autonomous environments discover one another, establish trust, negotiate rules, and execute open-ended tasks, cannot arise from simple combinations of P2P overlays and multi-agent systems. A layered architecture centered on a protocol adaptation layer is required to propagate semantic declarations about intentions, capabilities, states, and cooperation constraints. The architecture supports three core mechanisms: bodyless gossip with sequential logs for semantic announcement propagation, BAID-based identity binding with MG-EigenTrust for verifiable identity and

What carries the argument

protocol adaptation layer that connects upper-level task semantics with lower-level network operations

If this is right

  • Semantic announcement propagation via bodyless gossip enables collaborator discovery across open networks.
  • BAID-based identity binding combined with MG-EigenTrust supports verifiable governance of cooperation.
  • Semantic-gradient mechanism design allows open-ended task execution driven by attribution feedback.
  • The resulting framework supplies a system-level foundation for open, trustworthy, and scalable agent collaboration.

Where Pith is reading between the lines

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

  • If the architecture succeeds, agent collaboration could proceed without reliance on central servers, increasing resilience to single points of failure.
  • The mechanisms could be extended by testing discovery success rates in real edge-device deployments with varying participant counts.
  • This layered approach may change how LLM-based agents are orchestrated by requiring explicit semantic layers rather than direct overlay use.

Load-bearing premise

Heterogeneous agents on personal devices, edge nodes, or autonomous environments can and will discover one another, establish trust, and negotiate cooperation rules in an open peer-to-peer setting using the proposed mechanisms.

What would settle it

A deployment or simulation in which agents using the protocol adaptation layer and proposed mechanisms fail to discover collaborators or establish verifiable trust under realistic device heterogeneity, network churn, or adversarial conditions.

Figures

Figures reproduced from arXiv: 2606.17368 by Deen Ma, Shengli Zhang, Taotao Wang, Zibin Lin.

Figure 1
Figure 1. Figure 1: Reference architecture of a distributed general-purpose agent network. A node contains a general-purpose agent, a protocol adaptation layer, and a [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Main modules of the protocol adaptation layer. The layer links collaborator discovery, cooperation governance, and task execution into a feedback [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Bodyless-gossip-based semantic announcement propagation and performance-analysis framework. The protocol separates lightweight semantic-digest [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Discovery simulation trends under churn. Topic/OpenAgent maintains high success while reducing redundant traffic relative to public broadcast; [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: BAID identity binding workflow. The mechanism binds local agent code, user responsibility, and on-chain accountability records, and supports proof [PITH_FULL_IMAGE:figures/full_fig_p010_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Effect of recursive proof depth on BAID proof generation and verification overhead. The prototype compares AutoGPT, ReAct, and SmolAgents; [PITH_FULL_IMAGE:figures/full_fig_p011_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Effect of terminal payload size on BAID proof generation and verification overhead. Larger payloads increase proof generation time, while verification [PITH_FULL_IMAGE:figures/full_fig_p011_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Compact MG-EigenTrust simulation results redrawn in vector form. The panels summarize cross-topic attack exposure, burn-only attack ROI, targeted [PITH_FULL_IMAGE:figures/full_fig_p013_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Automated mechanism design for open agent collaboration. Mechanism-generation agents and strategy-exploration agents interact in a Stackelberg-style [PITH_FULL_IMAGE:figures/full_fig_p015_9.png] view at source ↗
read the original abstract

Large language models have accelerated the transition from passive conversational assistants to autonomous agents that can understand goals, plan actions, invoke tools, and execute multi-step tasks. Yet the capability of a single agent remains constrained by its local data, tool permissions, runtime environment, and governance boundary. This paper studies distributed general-purpose agent networks: open peer-to-peer networks in which heterogeneous agents deployed on personal devices, edge nodes, or autonomous computing environments can discover one another, establish trust, negotiate cooperation rules, and execute open-ended tasks. We argue that such networks cannot be obtained by simply combining existing peer-to-peer overlays with conventional multi-agent systems. Unlike traditional P2P networks, agent networks must propagate semantic declarations about intentions, capabilities, states, and cooperation constraints. We therefore propose a layered architecture centered on a protocol adaptation layer that connects upper-level task semantics with lower-level network operations. Based on this architecture, the paper identifies three core mechanism problems: semantic announcement propagation for collaborator discovery, verifiable identity and multi-topic reputation for cooperation governance, and semantic-gradient mechanism design for open task execution. For each problem, we present a technical route, including bodyless gossip with sequential logs, BAID-based identity binding with MG-EigenTrust reputation, and a Stackelberg-style mechanism-generation loop driven by semantic attribution feedback. We further report prototype overhead results for BAID-style tiered verification and mechanism-level simulations of MG-EigenTrust under cross-topic disguise-collusion attacks. The resulting framework provides a system-level foundation for open, trustworthy, and scalable agent collaboration.

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.

Referee Report

2 major / 2 minor

Summary. The paper proposes that open peer-to-peer networks of heterogeneous autonomous agents require a dedicated layered architecture with a protocol adaptation layer to propagate semantic declarations of intentions, capabilities, states, and constraints. It identifies three core mechanism problems—semantic announcement propagation, verifiable identity/reputation, and semantic-gradient mechanism design—and presents technical routes (bodyless gossip with sequential logs, BAID+MG-EigenTrust, Stackelberg-style loop) along with prototype overhead measurements for BAID and MG-EigenTrust simulations under cross-topic disguise-collusion attacks.

Significance. If the mechanisms prove effective, the work could supply a system-level foundation for scalable, trustworthy agent collaboration beyond single-agent limits. The inclusion of concrete prototype overhead data and targeted simulations provides a modest empirical anchor, though the absence of broader validation limits immediate impact.

major comments (2)
  1. [Abstract, §4] Abstract and mechanism sections: the central claim that the protocol adaptation layer plus the three mechanisms enable discovery, trust establishment, and negotiated cooperation among untrusted heterogeneous agents in open P2P settings is not load-bearing supported by evidence; only isolated BAID overhead on prototypes and MG-EigenTrust simulations under one attack model are reported, with no end-to-end results on announcement propagation, identity binding, or mechanism negotiation in topologies exhibiting churn, partial participation, or adversarial nodes.
  2. [§3] Architecture section: the assertion that agent networks cannot be obtained by combining existing P2P overlays with conventional multi-agent systems is presented without a concrete comparison, counter-example, or failure-mode analysis showing where standard approaches break on semantic propagation; this weakens the justification for introducing the protocol adaptation layer as a required component.
minor comments (2)
  1. [Introduction, §2] Notation for new constructs (bodyless gossip, BAID, MG-EigenTrust, semantic-gradient) is introduced without explicit cross-references to prior literature or precise formal definitions in the early sections.
  2. [Prototype evaluation] Prototype results for BAID tiered verification omit details on experimental configuration, number of runs, hardware, or baseline comparisons, making the overhead claims difficult to interpret.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on the strength of evidence and architectural justification. We address each major comment below.

read point-by-point responses

  1. Referee: [Abstract, §4] Abstract and mechanism sections: the central claim that the protocol adaptation layer plus the three mechanisms enable discovery, trust establishment, and negotiated cooperation among untrusted heterogeneous agents in open P2P settings is not load-bearing supported by evidence; only isolated BAID overhead on prototypes and MG-EigenTrust simulations under one attack model are reported, with no end-to-end results on announcement propagation, identity binding, or mechanism negotiation in topologies exhibiting churn, partial participation, or adversarial nodes.

    Authors: The reported results are limited to component-level overhead measurements for BAID and targeted MG-EigenTrust simulations under cross-topic disguise-collusion, as described in the manuscript. These provide initial feasibility evidence for the individual mechanisms but do not constitute end-to-end validation across dynamic topologies. We will revise the abstract and §4 to explicitly scope the current empirical contributions as mechanism-level validation and add a limitations subsection outlining the need for future end-to-end experiments under churn and partial participation. revision: yes

  2. Referee: [§3] Architecture section: the assertion that agent networks cannot be obtained by combining existing P2P overlays with conventional multi-agent systems is presented without a concrete comparison, counter-example, or failure-mode analysis showing where standard approaches break on semantic propagation; this weakens the justification for introducing the protocol adaptation layer as a required component.

    Authors: The manuscript's argument rests on the requirement for propagating semantic declarations of intentions, capabilities, states, and constraints, which differs from data-centric P2P or task-oriented MAS. To strengthen this, we will add a dedicated paragraph and table in §3 with concrete counter-examples (e.g., failure of standard gossip protocols to bind semantic constraints to identity without a dedicated layer, and limitations of FIPA-style ACL in open untrusted P2P settings) and failure-mode analysis for semantic propagation. revision: yes

Circularity Check

0 steps flagged

No circularity: architectural proposal with independent prototype measurements

full rationale

The paper advances an architectural argument and three mechanism proposals (bodyless gossip, BAID+MG-EigenTrust, Stackelberg loop) supported by overhead measurements on prototypes and targeted simulations. No equations, fitted parameters, or predictions appear that reduce by construction to the inputs. The central claim that a protocol adaptation layer is required is presented as a design rationale rather than a derived theorem; the reported results are direct measurements and attack-specific simulations, not self-referential fits. No load-bearing self-citations or uniqueness theorems are invoked in the provided text. The derivation chain is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only the abstract is available; no specific free parameters, axioms, or invented entities can be extracted or verified from the provided text.

pith-pipeline@v0.9.1-grok · 5818 in / 993 out tokens · 39728 ms · 2026-06-27T02:55:12.360449+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

52 extracted references · 3 linked inside Pith

  1. [1]

    Large language model based multi-agents: A survey of progress and challenges,

    T. Guoet al., “Large language model based multi-agents: A survey of progress and challenges,” arXiv preprint arXiv:2402.01680, 2024

    Pith/arXiv arXiv 2024

  2. [2]

    Cognitive architectures for language agents,

    T. R. Sumers, S. Yao, K. Narasimhan, and T. L. Griffiths, “Cognitive architectures for language agents,” arXiv preprint arXiv:2309.02427, 2023

    Pith/arXiv arXiv 2023

  3. [3]

    Big ideas 2026,

    ARK Investment Management LLC, “Big ideas 2026,” ARK Invest,

    2026

  4. [4]

    Available: https://ark-invest.com/big-ideas-2026

    [Online]. Available: https://ark-invest.com/big-ideas-2026

    2026

  5. [5]

    Openclaw ai chatbots are running amok – these scientists are listening in,

    M. Basu, “Openclaw ai chatbots are running amok – these scientists are listening in,”Nature, vol. 650, pp. 533–534, 2026

    2026

  6. [6]

    Internet of agents: Weaving a web of heterogeneous agents for collaborative intelligence,

    W. Chen, Z. You, R. Liet al., “Internet of agents: Weaving a web of heterogeneous agents for collaborative intelligence,”International Conference on Learning Representations, 2025

    2025

  7. [7]

    Internet of agents: Fundamentals, applications, and challenges,

    Y . Wang, S. Guo, Y . Panet al., “Internet of agents: Fundamentals, applications, and challenges,”IEEE Transactions on Cognitive Com- munications and Networking, 2025

    2025

  8. [8]

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

    Y . Yang, H. Chai, S. Shaoet al., “Agentnet: Decentralized evolutionary coordination for llm-based multi-agent systems,” inAdvances in Neural Information Processing Systems, 2025

    2025

  9. [9]

    Guiding opinions on deeply implementing the “artificial intelligence plus

    State Council of the People’s Republic of China, “Guiding opinions on deeply implementing the “artificial intelligence plus” action,” Guo Fa

  10. [10]

    11, Beijing, China, 2025

    No. 11, Beijing, China, 2025

    2025

  11. [11]

    Randomized rumor spreading,

    R. Karp, C. Schindelhauer, S. Shenker, and B. V ocking, “Randomized rumor spreading,” inProceedings of the IEEE Symposium on Founda- tions of Computer Science, 2000, pp. 565–574

    2000

  12. [12]

    Cyclon: Inexpensive membership management for unstructured p2p overlays,

    S. V oulgaris, D. Gavidia, and M. van Steen, “Cyclon: Inexpensive membership management for unstructured p2p overlays,”Journal of Network and Systems Management, vol. 13, no. 2, pp. 197–217, 2005

    2005

  13. [13]

    Gossip-based peer sampling,

    M. Jelasity, S. V oulgaris, R. Guerraoui, A.-M. Kermarrec, and M. van Steen, “Gossip-based peer sampling,”ACM Transactions on Computer Systems, vol. 25, no. 3, 2007

    2007

  14. [14]

    Gossipsub: Attack-resilient message propagation in the filecoin and eth2.0 networks,

    D. Vyzovitis, Y . Napora, D. McCormicket al., “Gossipsub: Attack-resilient message propagation in the filecoin and eth2.0 networks,” arXiv preprint arXiv:2007.02754, 2020. [Online]. Available: https://arxiv.org/abs/2007.02754

    arXiv 2007

  15. [15]

    Preamble and imreceiving for improved large message handling in libp2p gossipsub,

    M. U. Farooq and D. Kaiser, “Preamble and imreceiving for improved large message handling in libp2p gossipsub,” arXiv preprint arXiv:2505.17337, 2025. [Online]. Available: https://arxiv.org/abs/2505. 17337

    arXiv 2025

  16. [16]

    Bip152: Compact block relay,

    M. Corallo, “Bip152: Compact block relay,” Bitcoin Improvement Proposal, 2016. [Online]. Available: https://bips.dev/152

    2016

  17. [17]

    Bodyless block propagation: Tps fully scalable blockchain with pre-validation,

    C. Zhao, S. Zhang, T. Wanget al., “Bodyless block propagation: Tps fully scalable blockchain with pre-validation,”Future Generation Computer Systems, vol. 163, p. 107516, 2025

    2025

  18. [18]

    Certainsync: Rateless set reconciliation with certainty,

    T. Keniagin, E. Yaakobi, and O. Rottenstreich, “Certainsync: Rateless set reconciliation with certainty,”Proceedings of the ACM on Measurement and Analysis of Computing Systems, vol. 9, no. 2, pp. 1–33, 2025

    2025

  19. [19]

    Scaling blockchains with error correction codes: A survey on coded blockchains,

    C. Yang, K. W. Chin, J. Wanget al., “Scaling blockchains with error correction codes: A survey on coded blockchains,”ACM Computing Surveys, vol. 56, no. 6, pp. 1–33, 2024

    2024

  20. [20]

    Time, clocks, and the ordering of events in a distributed system,

    L. Lamport, “Time, clocks, and the ordering of events in a distributed system,”Communications of the ACM, vol. 21, no. 7, pp. 558–565, 1978

    1978

  21. [21]

    Timestamps in message-passing systems that preserve the partial ordering,

    C. J. Fidge, “Timestamps in message-passing systems that preserve the partial ordering,” inProceedings of the 11th Australian Computer Science Conference, 1988, pp. 56–66

    1988

  22. [22]

    Conflict-free replicated data types,

    M. Shapiro, N. Preguica, C. Baquero, and M. Zawirski, “Conflict-free replicated data types,”Stabilization, Safety, and Security of Distributed Systems, pp. 386–400, 2011. 17 TABLE VI BASELINES FOR EVALUATING AUTOMATED COOPERATION-MECHANISM DESIGN. Method Rule source Optimization mode Main limitation Manual mechanism Human-written rules Static design Diffi...

    2011

  23. [23]

    Practical byzantine fault tolerance,

    M. Castro and B. Liskov, “Practical byzantine fault tolerance,” in Proceedings of the Symposium on Operating Systems Design and Implementation, 1999, pp. 173–186

    1999

  24. [24]

    Hotstuff: Bft consensus with linearity and responsiveness,

    M. Yin, D. Malkhi, M. K. Reiteret al., “Hotstuff: Bft consensus with linearity and responsiveness,” inProceedings of the ACM Symposium on Principles of Distributed Computing, 2019, pp. 347–356

    2019

  25. [25]

    Announcing the agent2agent protocol (a2a),

    Google Developers Blog, “Announcing the agent2agent protocol (a2a),” 2025. [Online]. Available: https://developers.googleblog.com/en/ a2a-a-new-era-of-agent-interoperability

    2025

  26. [26]

    Agent name service (ans): A universal directory for secure ai agent discovery and interoperability,

    K. Huang and V . S. Narajala, “Agent name service (ans): A universal directory for secure ai agent discovery and interoperability,” arXiv preprint arXiv:2505.10609, 2025. [Online]. Available: https: //arxiv.org/abs/2505.10609

    arXiv 2025

  27. [27]

    The coral protocol: Open infrastructure connecting the internet of agents,

    R. J. Georgio, C. Forder, S. Debet al., “The coral protocol: Open infrastructure connecting the internet of agents,” arXiv preprint arXiv:2505.00749, 2025. [Online]. Available: https://arxiv.org/abs/2505. 00749

    arXiv 2025

  28. [28]

    The sybil attack,

    J. R. Douceur, “The sybil attack,” inProceedings of the International Workshop on Peer-to-Peer Systems, 2002, pp. 251–260

    2002

  29. [29]

    A survey of trust and reputation systems for online service provision,

    A. Jøsang, R. Ismail, and C. Boyd, “A survey of trust and reputation systems for online service provision,”Decision Support Systems, vol. 43, no. 2, pp. 618–644, 2007

    2007

  30. [30]

    Review on computational trust and reputation models,

    J. Sabater and C. Sierra, “Review on computational trust and reputation models,”Artificial Intelligence Review, vol. 24, no. 1, pp. 33–60, 2005

    2005

  31. [31]

    The eigentrust algorithm for reputation management in p2p networks,

    S. D. Kamvar, M. T. Schlosser, and H. Garcia-Molina, “The eigentrust algorithm for reputation management in p2p networks,” inProceedings of the International Conference on World Wide Web, 2003, pp. 640–651

    2003

  32. [32]

    Powertrust: A robust and scalable reputation system for trusted peer-to-peer computing,

    R. Zhou and K. Hwang, “Powertrust: A robust and scalable reputation system for trusted peer-to-peer computing,”IEEE Transactions on Parallel and Distributed Systems, vol. 18, no. 4, pp. 460–473, 2007

    2007

  33. [33]

    Identity management for agentic ai: The new frontier of authorization, authentication, and security for an ai agent world,

    T. South, S. Nagabhushanaradhya, A. Dissanayakaet al., “Identity management for agentic ai: The new frontier of authorization, authentication, and security for an ai agent world,” arXiv preprint arXiv:2510.25819, 2025. [Online]. Available: https://arxiv.org/abs/2510. 25819

    arXiv 2025

  34. [34]

    Ids for ai systems,

    A. Chan, N. Kolt, P. Willset al., “Ids for ai systems,” arXiv preprint arXiv:2406.12137, 2024. [Online]. Available: https://arxiv.org/abs/2406. 12137

    arXiv 2024

  35. [35]

    Beyond dns: Unlocking the internet of ai agents via the nanda index and verified agentfacts,

    R. Raskar, P. Chari, J. Zinkyet al., “Beyond dns: Unlocking the internet of ai agents via the nanda index and verified agentfacts,” arXiv preprint arXiv:2507.14263, 2025. [Online]. Available: https://arxiv.org/abs/2507.14263

    arXiv 2025

  36. [36]

    Loka protocol: A decentralized framework for trustworthy and ethical ai agent ecosystems,

    R. Ranjan, S. Gupta, and S. N. Singh, “Loka protocol: A decentralized framework for trustworthy and ethical ai agent ecosystems,” arXiv preprint arXiv:2504.10915, 2025. [Online]. Available: https://arxiv.org/ abs/2504.10915

    arXiv 2025

  37. [37]

    Erc-8004: Trustless agents,

    M. De Rossi, D. Crapis, J. Elliset al., “Erc-8004: Trustless agents,” Ethereum Improvement Proposal, 2025. [Online]. Available: https://eips.ethereum.org/EIPS/eip-8004

    2025

  38. [38]

    Reputation as a solution to cooperation collapse in llm-based mass,

    S. Ren, W. Fu, X. Zouet al., “Reputation as a solution to cooperation collapse in llm-based mass,” arXiv preprint arXiv:2505.05029, 2025. [Online]. Available: https://arxiv.org/abs/2505.05029

    arXiv 2025

  39. [39]

    Bottom-up reputation promotes cooperation with multi-agent reinforcement learning,

    T. Ren, X. Yao, Y . Liet al., “Bottom-up reputation promotes cooperation with multi-agent reinforcement learning,” inProceedings of the Interna- tional Conference on Autonomous Agents and Multiagent Systems, 2025, pp. 1745–1753

    2025

  40. [40]

    Talk, judge, cooperate: Gossip-driven indirect reciprocity in self-interested llm agents,

    S. Zhu, Y . Lin, S. Kaisthaet al., “Talk, judge, cooperate: Gossip-driven indirect reciprocity in self-interested llm agents,” arXiv preprint arXiv:2602.07777, 2026. [Online]. Available: https: //arxiv.org/abs/2602.07777

    Pith/arXiv arXiv 2026

  41. [41]

    Drf: Llm-agent dynamic reputation filtering framework,

    Y . Lou, H. Hu, S. Maet al., “Drf: Llm-agent dynamic reputation filtering framework,” inProceedings of the International Conference on Neural Information Processing, 2025, pp. 127–141

    2025

  42. [42]

    Complexity of mechanism design,

    V . Conitzer and T. Sandholm, “Complexity of mechanism design,” Proceedings of the Conference on Uncertainty in Artificial Intelligence, pp. 103–110, 2002

    2002

  43. [43]

    Automated mechanism design: A new application area for search algorithms,

    T. Sandholm, “Automated mechanism design: A new application area for search algorithms,” inProceedings of the International Joint Conference on Artificial Intelligence, 2003, pp. 90–97

    2003

  44. [44]

    Automated mechanism design: A survey,

    M. J. Curry, Z. Fan, Y . Jianget al., “Automated mechanism design: A survey,”ACM SIGecom Exchanges, vol. 22, no. 2, pp. 102–120, 2025

    2025

  45. [45]

    Optimal auctions through deep learning,

    P. Duetting, Z. Feng, H. Narasimhan, D. C. Parkes, and S. S. Ravin- dranath, “Optimal auctions through deep learning,” inProceedings of the International Conference on Machine Learning, 2019, pp. 1706–1715

    2019

  46. [46]

    Deep automated mechanism design for integrating ad auction and allocation in feed,

    X. Li, Z. Wang, B. Zhuet al., “Deep automated mechanism design for integrating ad auction and allocation in feed,” inProceedings of the International ACM SIGIR Conference on Research and Development in Information Retrieval, 2024, pp. 1211–1220

    2024

  47. [47]

    Edge-awarenet: A graph neural approach for efficient multi-item auction mechanism design,

    X. Cao, Y . Yang, J. Liet al., “Edge-awarenet: A graph neural approach for efficient multi-item auction mechanism design,”Knowledge-Based Systems, vol. 313, p. 115097, 2025

    2025

  48. [48]

    Research and applications of game intelligence,

    J. Hao, K. Shao, K. Liet al., “Research and applications of game intelligence,”Scientia Sinica Informationis, vol. 53, no. 10, pp. 1892– 1923, 2023, in Chinese

    1923

  49. [49]

    Robust reinforcement learning as a stackelberg game via adaptively-regularized adversarial training,

    P. Huang, M. Xu, F. Fanget al., “Robust reinforcement learning as a stackelberg game via adaptively-regularized adversarial training,” inPro- ceedings of the International Joint Conference on Artificial Intelligence, 2022, pp. 3099–3106

    2022

  50. [50]

    Game theory meets large language models: A systematic survey,

    H. Sun, Y . Wu, Y . Chenget al., “Game theory meets large language models: A systematic survey,” inProceedings of the International Joint Conference on Artificial Intelligence, 2025, pp. 10 669–10 677

    2025

  51. [51]

    Optimizing generative ai by backpropagating language model feedback,

    M. Yuksekgonul, F. Bianchi, J. Boenet al., “Optimizing generative ai by backpropagating language model feedback,”Nature, vol. 639, pp. 609–616, 2025

    2025

  52. [52]

    Generative agents: Interactive simulacra of human behavior,

    J. S. Park, J. O’Brien, C. J. Cai, M. R. Morris, P. Liang, and M. S. Bernstein, “Generative agents: Interactive simulacra of human behavior,” inProceedings of the ACM Symposium on User Interface Software and Technology, 2023, pp. 1–22

    2023