arxiv: 2604.02334 · v1 · submitted 2026-01-18 · 💻 cs.AI · cs.MA

Recognition: 1 theorem link

· Lean Theorem

Holos: A Web-Scale LLM-Based Multi-Agent System for the Agentic Web

Xiaohang Nie , Zihan Guo , Zicai Cui , Jiachi Yang , Zeyi Chen , Leheyi De , Yu Zhang , Junwei Liao

show 15 more authors

Bo Huang Yingxuan Yang Zhi Han Zimian Peng Linyao Chen Wenzheng Tom Tang Zongkai Liu Tao Zhou Botao Amber Hu Shuyang Tang Jianghao Lin Weiwen Liu Muning Wen Yuanjian Zhou Weinan Zhang

Authors on Pith no claims yet

Pith reviewed 2026-05-16 13:15 UTC · model grok-4.3

classification 💻 cs.AI cs.MA

keywords LLM agentsmulti-agent systemsAgentic Webfive-layer architectureNuwa enginemarket orchestratorvalue cycleagent persistence

0 comments

The pith

Holos presents a five-layer architecture using the Nuwa engine, a market-driven Orchestrator, and an endogenous value cycle to enable scalable and persistent LLM-based multi-agent systems in the Agentic Web.

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

The paper is trying to establish that a carefully designed five-layer system can solve the practical problems preventing LLM agents from forming a large, self-sustaining web. It describes Holos as incorporating an engine for quick agent creation, market rules for coordination, and internal value flows to keep agents motivated without external rewards. A sympathetic reader would care if this works because it points to a path where agents can operate at internet scale, interact autonomously, and evolve over time rather than staying limited to single tasks. The release of the system online invites testing whether such an architecture actually produces the promised coordination and persistence.

Core claim

Holos adopts a five-layer architecture, with core modules primarily featuring the Nuwa engine for high-efficiency agent generation and hosting, a market-driven Orchestrator for resilient coordination, and an endogenous value cycle to achieve incentive compatibility. By bridging the gap between micro-level collaboration and macro-scale emergence, Holos hopes to lay the foundation for the next generation of the self-organizing and continuously evolving Agentic Web.

What carries the argument

Five-layer architecture with the Nuwa engine for agent generation and hosting, market-driven Orchestrator for coordination, and endogenous value cycle for incentive compatibility.

If this is right

Agents can be generated and hosted efficiently at web scale.
Coordination remains resilient through market-driven mechanisms.
Incentive compatibility is maintained via endogenous value cycles.
Micro-level agent interactions can lead to macro-scale emergent behaviors.
The system supports long-term ecological persistence of the agent ecosystem.

Where Pith is reading between the lines

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

This architecture might enable truly open-ended agent evolution similar to biological ecosystems.
It could be extended to incorporate real economic incentives from external markets.
The public release allows empirical validation of coordination efficiency at increasing scales.
Similar layered designs could be applied to other domains like robotic swarms or distributed computing.

Load-bearing premise

The described five-layer architecture, particularly the endogenous value cycle and market-driven Orchestrator, will successfully bridge micro-level collaboration to macro-scale emergence and achieve long-term ecological persistence in an open-world setting.

What would settle it

Running Holos with thousands of agents over weeks and measuring if coordination failures or value dissipation occur, or if stable higher-order structures emerge.

Figures

Figures reproduced from arXiv: 2604.02334 by Bo Huang, Botao Amber Hu, Jiachi Yang, Jianghao Lin, Junwei Liao, Leheyi De, Linyao Chen, Muning Wen, Shuyang Tang, Tao Zhou, Weinan Zhang, Weiwen Liu, Wenzheng Tom Tang, Xiaohang Nie, Yingxuan Yang, Yuanjian Zhou, Yu Zhang, Zeyi Chen, Zhi Han, Zicai Cui, Zihan Guo, Zimian Peng, Zongkai Liu.

**Figure 1.** Figure 1: Screenshot of Holos’ user interface. a is the user entry on the homepage, where users can submit requests via natural language; b is the directed acyclic graph (DAG) of the Orchestrator during task execution; c shows details of a subtask, including mission responsibility, results, and evaluation; d displays agent outputs and an interactive workspace; e visualizes all agents as a galaxy and highlights selec… view at source ↗

**Figure 2.** Figure 2: The five-layer architecture of Holos. This architecture orchestrates the end-to-end lifecycle [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 3.** Figure 3: Results of the scale-invariant discovery efficiency test. The left subgraph displays the [PITH_FULL_IMAGE:figures/full_fig_p019_3.png] view at source ↗

**Figure 4.** Figure 4: Results for adaptive self-healing resilience under varying failure injection probabilities ( [PITH_FULL_IMAGE:figures/full_fig_p019_4.png] view at source ↗

**Figure 5.** Figure 5: Analysis of resource allocation effectiveness and ability identification efficiency of Holos. [PITH_FULL_IMAGE:figures/full_fig_p020_5.png] view at source ↗

**Figure 6.** Figure 6: Evolutionary resilience and market selection evaluation of the economic system in Holos. [PITH_FULL_IMAGE:figures/full_fig_p021_6.png] view at source ↗

**Figure 7.** Figure 7: Adaptation under dynamic market entry of the economic system in Holos. The left [PITH_FULL_IMAGE:figures/full_fig_p022_7.png] view at source ↗

**Figure 8.** Figure 8: Demonstration of Holos versatility across diverse cases. (a) Geospatial Planning: Au [PITH_FULL_IMAGE:figures/full_fig_p023_8.png] view at source ↗

read the original abstract

As large language models (LLM)-driven agents transition from isolated task solvers to persistent digital entities, the emergence of the Agentic Web, an ecosystem where heterogeneous agents autonomously interact and co-evolve, marks a pivotal shift toward Artificial General Intelligence (AGI). However, LLM-based multi-agent systems (LaMAS) are hindered by open-world issues such as scaling friction, coordination breakdown, and value dissipation. To address these challenges, we introduce Holos, a web-scale LaMAS architected for long-term ecological persistence. Holos adopts a five-layer architecture, with core modules primarily featuring the Nuwa engine for high-efficiency agent generation and hosting, a market-driven Orchestrator for resilient coordination, and an endogenous value cycle to achieve incentive compatibility. By bridging the gap between micro-level collaboration and macro-scale emergence, Holos hopes to lay the foundation for the next generation of the self-organizing and continuously evolving Agentic Web. We have publicly released Holos (accessible at https://holosai.io), providing a resource for the community and a testbed for future research in large-scale agentic ecosystems.

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

Holos is a high-level five-layer architecture proposal for web-scale LLM multi-agent systems that names useful components but supplies no experiments or data to show they work.

read the letter

Holos lays out a five-layer setup for running many LLM agents together on the web, with the Nuwa engine handling fast agent creation and hosting, a market orchestrator for coordination, and an endogenous value cycle meant to keep incentives aligned without outside rewards. The authors have put the system online at holosai.io, which lets others inspect or extend it directly. That release is the most concrete thing here. The paper does a clear job naming the real headaches in current LaMAS work: agents that scale poorly, lose coordination as numbers grow, and stop participating once value leaks away. Framing the fix around a persistent, self-evolving Agentic Web gives the proposal a coherent target. The market and value-cycle ideas are reasonable extensions of existing incentive work in multi-agent systems. The main weakness is the total lack of supporting evidence. There are no simulations, no ablation runs, no metrics on coordination stability or participation rates, and no checks on whether the market mechanism actually produces stable equilibria under noisy LLM behavior. The claims about bridging micro-level agent talks to macro-scale emergence rest on the architecture description alone. Without those tests it is impossible to judge if the design delivers on its promises or just restates the problems in new terms. This is for readers already deep in multi-agent LLM research who want architectural sketches and a public codebase to play with. It is not yet useful for anyone needing validated results or reproducible methods. The work shows clear thinking about the problem space and engages honestly with the scaling issues, so it deserves peer review rather than a desk reject. Referees will almost certainly push for experiments, but the public release gives a base to build those on.

Referee Report

2 major / 1 minor

Summary. The manuscript proposes Holos, a web-scale LLM-based multi-agent system (LaMAS) for the Agentic Web. It introduces a five-layer architecture whose core modules are the Nuwa engine for high-efficiency agent generation and hosting, a market-driven Orchestrator for resilient coordination, and an endogenous value cycle intended to achieve incentive compatibility. The central claim is that this design bridges micro-level agent collaboration to macro-scale emergence while mitigating scaling friction, coordination breakdown, and value dissipation; a public release at https://holosai.io is provided as a community testbed.

Significance. If the architecture can be shown through evaluation to produce stable coordination and sustained participation, Holos would offer a concrete, deployable framework for persistent, incentive-aligned agent ecosystems at web scale. The public release would then serve as a valuable shared resource for studying self-organizing multi-agent dynamics, potentially shaping subsequent work on open-world LaMAS and long-term ecological persistence in agentic systems.

major comments (2)

[Abstract] Abstract: The claims that the five-layer architecture, market-driven Orchestrator, and endogenous value cycle successfully bridge micro-level collaboration to macro-scale emergence and solve scaling friction, coordination breakdown, and value dissipation rest entirely on descriptive design intuition. No experiments, simulations, ablation studies, equilibrium analysis, or benchmarks are supplied to demonstrate that the market mechanism yields stable equilibria or that the value cycle sustains participation under realistic LLM noise and heterogeneous capabilities.
[Architecture] Architecture section (five-layer stack): The description of the Nuwa engine and endogenous value cycle provides high-level module outlines but contains no quantitative metrics, parameter settings, or performance comparisons. Without such data it is impossible to assess whether the claimed efficiency gains or incentive compatibility are realized or even measurable.

minor comments (1)

The manuscript would benefit from an explicit evaluation roadmap or preliminary results section to allow readers to judge the feasibility of the proposed mechanisms.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. The work presents Holos as an architectural proposal for web-scale LaMAS, supported by a public deployment at https://holosai.io. We address the major comments point by point below, acknowledging the primarily descriptive nature of the current version.

read point-by-point responses

Referee: [Abstract] The claims that the five-layer architecture, market-driven Orchestrator, and endogenous value cycle successfully bridge micro-level collaboration to macro-scale emergence and solve scaling friction, coordination breakdown, and value dissipation rest entirely on descriptive design intuition. No experiments, simulations, ablation studies, equilibrium analysis, or benchmarks are supplied to demonstrate that the market mechanism yields stable equilibria or that the value cycle sustains participation under realistic LLM noise and heterogeneous capabilities.

Authors: We agree that the abstract and manuscript rely on design rationale without empirical validation. As a conceptual framework paper, the focus is on proposing the architecture and releasing the system as a community testbed. In revision we will add a dedicated section discussing initial observations from https://holosai.io, preliminary coordination metrics, and an outline of planned equilibrium analysis and simulations under noisy LLM conditions. revision: partial
Referee: [Architecture] Architecture section (five-layer stack): The description of the Nuwa engine and endogenous value cycle provides high-level module outlines but contains no quantitative metrics, parameter settings, or performance comparisons. Without such data it is impossible to assess whether the claimed efficiency gains or incentive compatibility are realized or even measurable.

Authors: The architecture section intentionally emphasizes conceptual design. Implementation-level details, including parameter settings for the Nuwa engine and value-cycle mechanics, are provided in the linked open-source release. We will revise the section to include concrete example configurations, initial runtime metrics from the deployed system, and a comparison table of generation efficiency. revision: yes

Circularity Check

0 steps flagged

No circularity: purely descriptive architecture with no derivations, equations, or self-referential reductions.

full rationale

The manuscript introduces Holos via a five-layer architecture, Nuwa engine, market-driven Orchestrator, and endogenous value cycle, but supplies only high-level descriptive claims without any equations, parameter fittings, derivations, or load-bearing self-citations. No step reduces a claimed outcome to its own inputs by construction, nor renames a fitted result as a prediction. The bridging claim from micro to macro emergence is asserted as design intent rather than derived, leaving the paper self-contained as an architectural proposal. This matches the default non-circular case for descriptive system papers.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 3 invented entities

The central claims rest on domain assumptions about agent behavior and ad-hoc architectural inventions without independent evidence or derivations provided.

axioms (2)

domain assumption LLM-driven agents can transition from isolated solvers to persistent digital entities that autonomously interact and co-evolve in an Agentic Web ecosystem
Stated directly in the opening of the abstract as the emergence marking a shift toward AGI.
ad hoc to paper The five-layer architecture with Nuwa engine, market-driven Orchestrator, and endogenous value cycle can bridge micro-level collaboration and macro-scale emergence while solving scaling friction, coordination breakdown, and value dissipation
Core architectural claim presented without supporting derivation or data.

invented entities (3)

Nuwa engine no independent evidence
purpose: High-efficiency agent generation and hosting
New core module introduced for the architecture.
market-driven Orchestrator no independent evidence
purpose: Resilient coordination of heterogeneous agents
New coordination component proposed to prevent breakdown.
endogenous value cycle no independent evidence
purpose: Achieve incentive compatibility and prevent value dissipation
Invented mechanism for long-term persistence.

pith-pipeline@v0.9.0 · 5580 in / 1503 out tokens · 60271 ms · 2026-05-16T13:15:32.980243+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

?

unclear
Relation between the paper passage and the cited Recognition theorem.

Holos adopts a five-layer architecture, with core modules primarily featuring the Nuwa engine for high-efficiency agent generation and hosting, a market-driven Orchestrator for resilient coordination, and an endogenous value cycle to achieve incentive compatibility.

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Forward citations

Cited by 1 Pith paper

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

Externalization in LLM Agents: A Unified Review of Memory, Skills, Protocols and Harness Engineering
cs.SE 2026-04 accept novelty 5.0

LLM agent progress depends on externalizing cognitive functions into memory, skills, protocols, and harness engineering that coordinates them reliably.

Reference graph

Works this paper leans on

103 extracted references · 103 canonical work pages · cited by 1 Pith paper · 9 internal anchors

[1]

Relational inductive biases, deep learning, and graph networks

Peter W Battaglia, Jessica B Hamrick, Victor Bapst, Alvaro Sanchez-Gonzalez, Vinicius Zam- baldi, Mateusz Malinowski, Andrea Tacchetti, David Raposo, Adam Santoro, Ryan Faulkner, et al. Relational inductive biases, deep learning, and graph networks.arXiv preprint arXiv:1806.01261, 2018

work page internal anchor Pith review Pith/arXiv arXiv 2018
[2]

Wolpert and W.G

D.H. Wolpert and W.G. Macready. No free lunch theorems for optimization.IEEE Transactions on Evolutionary Computation, 1(1):67–82, 1997. doi: 10.1109/4235.585893

work page doi:10.1109/4235.585893 1997
[3]

No free lunch theorem: A review.Approximation and optimization: Algorithms, complexity and applications, pages 57–82, 2019

Stavros P Adam, Stamatios-Aggelos N Alexandropoulos, Panos M Pardalos, and Michael N Vrahatis. No free lunch theorem: A review.Approximation and optimization: Algorithms, complexity and applications, pages 57–82, 2019

work page 2019
[4]

MIT Press, Cambridge, MA, 2015

Thomas W Malone and Michael S Bernstein, editors.Handbook of Collective Intelligence. MIT Press, Cambridge, MA, 2015

work page 2015
[5]

Collective intelligence for deep learning: A survey of recent devel- opments.Collective Intelligence, 1(1):26339137221114874, 2022

David Ha and Yujin Tang. Collective intelligence for deep learning: A survey of recent devel- opments.Collective Intelligence, 1(1):26339137221114874, 2022

work page 2022
[6]

Magent: A many-agent reinforcement learning platform for artificial collective intelligence

Lianmin Zheng, Jiacheng Yang, Han Cai, Ming Zhou, Weinan Zhang, Jun Wang, and Yong Yu. Magent: A many-agent reinforcement learning platform for artificial collective intelligence. In Proceedings of the AAAI conference on artificial intelligence, volume 32, 2018

work page 2018
[7]

Penguin, 2017

Kevin Kelly.The inevitable: Understanding the 12 technological forces that will shape our future. Penguin, 2017

work page 2017
[8]

LLM-based multi- agent systems: Techniques and business perspectives, 2024

Yingxuan Yang, Qiuying Peng, Jun Wang, Ying Wen, and Weinan Zhang. LLM-based multi- agent systems: Techniques and business perspectives, 2024. URLhttps://arxiv.org/abs/24 11.14033

work page 2024
[9]

Agentic web: Weaving the next web with AI agents, 2025

Yingxuan Yang, Mulei Ma, Yuxuan Huang, Huacan Chai, Chenyu Gong, Haoran Geng, Yuan- jian Zhou, Ying Wen, Meng Fang, Muhao Chen, Shangding Gu, Ming Jin, Costas Spanos, Yang Yang, Pieter Abbeel, Dawn Song, Weinan Zhang, and Jun Wang. Agentic web: Weaving the next web with AI agents, 2025

work page 2025
[10]

A survey on large language model based autonomous agents.Frontiers of Computer Science, 18(6):186345, 2024

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

work page 2024
[11]

GPT-4 Technical Report

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

work page internal anchor Pith review Pith/arXiv arXiv 2023
[12]

Claude docs.https://platform.claude.com/docs/en/home, 2026

Claude. Claude docs.https://platform.claude.com/docs/en/home, 2026

work page 2026
[13]

AutoGPT.https://github.com/Significant-Gravitas/AutoGPT, 2023

Significant Gravitas. AutoGPT.https://github.com/Significant-Gravitas/AutoGPT, 2023

work page 2023
[14]

LangChain.https://github.com/langchain-ai/langchain, 2022

LangChain AI. LangChain.https://github.com/langchain-ai/langchain, 2022

work page 2022
[15]

MetaGPT: Meta Programming for A Multi-Agent Collaborative Framework

Sirui Hong, Mingchen Zhuge, Jiaqi Chen, Xiawu Zheng, Yuheng Cheng, Ceyao Zhang, Jinlin Wang, Zili Wang, Steven Ka Shing Yau, Zijuan Lin, Liyang Zhou, Chenyu Ran, Lingfeng Xiao, Chenglin Wu, and Jürgen Schmidhuber. MetaGPT: Meta programming for a multi-agent collaborative framework, 2024. URLhttps://arxiv.org/abs/2308.00352

work page internal anchor Pith review Pith/arXiv arXiv 2024
[16]

Chatdev: Communicative agents for software development

Chen Qian, Wei Liu, Hongzhang Liu, Nuo Chen, Yufan Dang, Jiahao Li, Cheng Yang, Weize Chen, Yusheng Su, Xin Cong, et al. Chatdev: Communicative agents for software development. InProceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 15174–15186, 2024. 32 Holos

work page 2024
[17]

Swarm.https://github.com/openai/swarm, 2024

OpenAI. Swarm.https://github.com/openai/swarm, 2024

work page 2024
[18]

AutoGen: Enabling next-gen LLM applications via multi-agent conversation,

Qingyun Wu, Gagan Bansal, Jieyu Zhang, Yiran Wu, Beibin Li, Erkang Zhu, Li Jiang, Xiaoyun Zhang, Shaokun Zhang, Jiale Liu, Ahmed Hassan Awadallah, Ryen W White, Doug Burger, and Chi Wang. AutoGen: Enabling next-gen LLM applications via multi-agent conversation,

work page
[19]

URLhttps://arxiv.org/abs/2308.08155

work page internal anchor Pith review Pith/arXiv arXiv
[20]

Architecting the Internet of AI Agents.https://projectnanda.org/#/, 2025

NANDA Community. Architecting the Internet of AI Agents.https://projectnanda.org/#/, 2025

work page 2025
[21]

Scaling large language model-based multi-agent collabo- ration.arXiv preprint arXiv:2406.07155, 2024

Chen Qian, Zihao Xie, Yifei Wang, Wei Liu, Kunlun Zhu, Hanchen Xia, Yufan Dang, Zhuoyun Du, Weize Chen, Cheng Yang, et al. Scaling large language model-based multi-agent collabo- ration.arXiv preprint arXiv:2406.07155, 2024

work page arXiv 2024
[22]

Towards a Science of Scaling Agent Systems

Yubin Kim, Ken Gu, Chanwoo Park, Chunjong Park, Samuel Schmidgall, A Ali Heydari, Yao Yan, Zhihan Zhang, Yuchen Zhuang, Mark Malhotra, et al. Towards a science of scaling agent systems.arXiv preprint arXiv:2512.08296, 2025

work page internal anchor Pith review Pith/arXiv arXiv 2025
[23]

The agentic economy.arXiv preprint arXiv:2505.15799, 2025

David M Rothschild, Markus Mobius, Jake M Hofman, Eleanor W Dillon, Daniel G Goldstein, Nicole Immorlica, Sonia Jaffe, Brendan Lucier, Aleksandrs Slivkins, and Matthew Vogel. The agentic economy.arXiv preprint arXiv:2505.15799, 2025

work page arXiv 2025
[24]

Agent2Agent (A2A) Protocol.https://a2a-protocol.org/latest, 2025

Google. Agent2Agent (A2A) Protocol.https://a2a-protocol.org/latest, 2025

work page 2025
[25]

ReAct: Synergizing reasoning and acting in language models

Shunyu Yao, Jeffrey Zhao, Dian Yu, Izhak Shafran, Karthik R Narasimhan, and Yuan Cao. ReAct: Synergizing reasoning and acting in language models. InNeurIPS 2022 Foundation Models for Decision Making Workshop, 2022. URLhttps://openreview.net/forum?id=tvI4 u1ylcqs

work page 2022
[26]

MCPZoo: A large-scale dataset of runnable model context protocol servers for AI agent.arXiv preprint arXiv:2512.15144, 2025

Mengying Wu, Pei Chen, Geng Hong, Aichao An, Jinsong Chen, Binwang Wan, Xudong Pan, Jiarun Dai, and Min Yang. MCPZoo: A large-scale dataset of runnable model context protocol servers for AI agent.arXiv preprint arXiv:2512.15144, 2025

work page arXiv 2025
[27]

A. B. Kahn. Topological sorting of large networks.Commun. ACM, 5(11):558–562, November

work page
[28]

doi: 10.1145/368996.369025

ISSN 0001-0782. doi: 10.1145/368996.369025. URLhttps://doi.org/10.1145/368996 .369025

work page doi:10.1145/368996.369025
[29]

From ranknet to lambdarank to lambdamart: An overview.Learning, 11(23-581):81, 2010

Christopher JC Burges. From ranknet to lambdarank to lambdamart: An overview.Learning, 11(23-581):81, 2010

work page 2010
[30]

An agent ecosystem with task-driven hierarchical evolving reasoners catalyzing proof of intelligence, 2025

Shuyang Tang, Zihan Guo, Yuanjian Zhou, Chenyi Wang, Linlin You, Minjie Bian, and Weinan Zhang. An agent ecosystem with task-driven hierarchical evolving reasoners catalyzing proof of intelligence, 2025. Available at SSRN:https://ssrn.com/abstract=5789723

work page 2025
[31]

Algorithm AS 136: A k-means clustering algorithm

John A Hartigan and Manchek A Wong. Algorithm AS 136: A k-means clustering algorithm. Journal of the royal statistical society. series c (applied statistics), 28(1):100–108, 1979

work page 1979
[32]

Demystifying evals for ai agents.https://www.anthropic.com/engineering/de mystifying-evals-for-ai-agents, 2026

Anthropic. Demystifying evals for ai agents.https://www.anthropic.com/engineering/de mystifying-evals-for-ai-agents, 2026

work page 2026
[33]

The sybil attack

John R Douceur. The sybil attack. InInternational workshop on peer-to-peer systems, pages 251–260. Springer, 2002

work page 2002
[34]

Sybilproof reputation mechanisms

Alice Cheng and Eric Friedman. Sybilproof reputation mechanisms. InProceedings of the 2005 ACM SIGCOMM workshop on Economics of peer-to-peer systems, pages 128–132, 2005

work page 2005
[35]

Generative agents: Interactive simulacra of human behavior

Joon Sung Park, Joseph O’Brien, Carrie Jun Cai, Meredith Ringel Morris, Percy Liang, and Michael S Bernstein. Generative agents: Interactive simulacra of human behavior. InProceed- ings of the 36th annual acm symposium on user interface software and technology, pages 1–22, 2023. 33 Holos

work page 2023
[36]

SoDA: An efficient interaction paradigm for the agentic web.arXiv preprint arXiv:2512.22135, 2025

Zicai Cui, Zhouyuan Jian, Weiwen Liu, and Weinan Zhang. SoDA: An efficient interaction paradigm for the agentic web.arXiv preprint arXiv:2512.22135, 2025

work page arXiv 2025
[37]

Theory of mind may have spontaneously emerged in large language models

Michal Kosinski. Theory of mind may have spontaneously emerged in large language models. arXiv preprint arXiv:2302.02083, 4:169, 2023

work page arXiv 2023
[38]

The rise and potential of large language model based agents: A survey.Science China Information Sciences, 68(2):121101, 2025

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

work page 2025
[39]

Kobayashi, and Weinan Zhang

Linyao Chen, Zimian Peng, Yingxuan Yang, Yikun Wang, Wenzheng Tom Tang, Hiroki H. Kobayashi, and Weinan Zhang. EnvX: Agentize everything with agentic AI, 2025. URLhttps: //arxiv.org/abs/2509.08088

work page arXiv 2025
[40]

Automated creation and enrichment framework for improved invocation of enterprise APIs as tools, 2025

Prerna Agarwal, Himanshu Gupta, Soujanya Soni, Rohith Vallam, Renuka Sindhgatta, and Sameep Mehta. Automated creation and enrichment framework for improved invocation of enterprise APIs as tools, 2025. URLhttps://arxiv.org/abs/2509.11626

work page arXiv 2025
[41]

Memory in the age of AI agents,

Yuyang Hu, Shichun Liu, Yanwei Yue, Guibin Zhang, Boyang Liu, Fangyi Zhu, Jiahang Lin, Honglin Guo, Shihan Dou, Zhiheng Xi, Senjie Jin, Jiejun Tan, Yanbin Yin, Jiongnan Liu, Zeyu Zhang, Zhongxiang Sun, Yutao Zhu, Hao Sun, Boci Peng, Zhenrong Cheng, Xuanbo Fan, Jiaxin Guo, Xinlei Yu, Zhenhong Zhou, Zewen Hu, Jiahao Huo, Junhao Wang, Yuwei Niu, Yu Wang, Zhe...

work page
[42]

URLhttps://arxiv.org/abs/2512.13564

work page internal anchor Pith review Pith/arXiv arXiv
[43]

Memp: Exploring Agent Procedural Memory

Runnan Fang, Yuan Liang, Xiaobin Wang, Jialong Wu, Shuofei Qiao, Pengjun Xie, Fei Huang, Huajun Chen, and Ningyu Zhang. Memp: Exploring agent procedural memory, 2025. URL https://arxiv.org/abs/2508.06433

work page internal anchor Pith review Pith/arXiv arXiv 2025
[44]

MemRL: Self- evolving agents via runtime reinforcement learning on episodic memory, 2026

Shengtao Zhang, Jiaqian Wang, Ruiwen Zhou, Junwei Liao, Yuchen Feng, Weinan Zhang, Ying Wen, Zhiyu Li, Feiyu Xiong, Yutao Qi, Bo Tang, and Muning Wen. MemRL: Self- evolving agents via runtime reinforcement learning on episodic memory, 2026. URLhttps: //arxiv.org/abs/2601.03192

work page arXiv 2026
[45]

Narwhal and Tusk: a DAG-based mempool and efficient BFT consensus

George Danezis, Lefteris Kokoris-Kogias, Alberto Sonnino, and Alexander Spiegelman. Narwhal and Tusk: a DAG-based mempool and efficient BFT consensus. InProceedings of the European Conference on Computer Systems, pages 34–50, 2022

work page 2022
[46]

SoK: DAG-based blockchain systems

Qin Wang, Jiangshan Yu, Shiping Chen, and Yang Xiang. SoK: DAG-based blockchain systems. ACM Comput. Surv., 55(12):261:1–261:38, 2023

work page 2023
[47]

The Bitcoin Lightning network: Scalable off-chain instant payments, version 0.5.9.2.https://lightning.network/lightning-network-paper.pdf,

Joseph Poon and Thaddeus Dryja. The Bitcoin Lightning network: Scalable off-chain instant payments, version 0.5.9.2.https://lightning.network/lightning-network-paper.pdf,

work page
[48]

18, 2026

Last access: Jan. 18, 2026

work page 2026
[49]

SoK: Layer-two blockchain protocols

Lewis Gudgeon, Pedro Moreno-Sanchez, Stefanie Roos, Patrick McCorry, and Arthur Gervais. SoK: Layer-two blockchain protocols. InFC, pages 201–226, 2020

work page 2020
[50]

Settling payments fast and private: Efficient decentralized routing for path-based transactions

Stefanie Roos, Pedro Moreno-Sanchez, Aniket Kate, and Ian Goldberg. Settling payments fast and private: Efficient decentralized routing for path-based transactions. InProceedings of the Network and Distributed System Security Symposium, 2018

work page 2018
[51]

Flash: Efficient dynamic routing for offchain networks

Peng Wang, Hong Xu, Xin Jin, and Tao Wang. Flash: Efficient dynamic routing for offchain networks. InProceedings of the ACM International Conference on Emerging Networking Ex- periments and Technologies, pages 370–381, 2019. 34 Holos

work page 2019
[52]

Anonymous multi-hop locks for blockchain scalability and interoperability

GiulioMalavolta, PedroMoreno-Sanchez, ClaraSchneidewind, AniketKate, andMatteoMaffei. Anonymous multi-hop locks for blockchain scalability and interoperability. InProceedings of the Network and Distributed System Security Symposium, 2019

work page 2019
[53]

Shuyang Tang and Sherman S. M. Chow. Strengthening multi-hop channels via strategic mesh connections. InProceedings of the Financial Cryptography and Data Security Conference, 2025

work page 2025
[54]

A pragmatic introduction to secure multi-party computation.Found

David Evans, Vladimir Kolesnikov, and Mike Rosulek. A pragmatic introduction to secure multi-party computation.Found. Trends Priv. Secur., 2(2-3):70–246, 2018

work page 2018
[55]

FRIDA:dataavailabilitysampling from FRI

MathiasHall-Andersen, MarkSimkin, andBenediktWagner. FRIDA:dataavailabilitysampling from FRI. InProceedings of the International Cryptology Conference, Part VI, pages 289–324, 2024

work page 2024
[56]

MD- ML: super fast privacy-preserving machine learning for malicious security with a dishonest majority

Boshi Yuan, Shixuan Yang, Yongxiang Zhang, Ning Ding, Dawu Gu, and Shi-Feng Sun. MD- ML: super fast privacy-preserving machine learning for malicious security with a dishonest majority. InUSENIX Security Symposium, 2024

work page 2024
[57]

SHAFT:secure, handy, accurateandfasttransformer inference

AndesY.L.KeiandShermanS.M.Chow. SHAFT:secure, handy, accurateandfasttransformer inference. InProceedings of the Network and Distributed System Security Symposium, 2025

work page 2025
[58]

Mosformer: Maliciously secure three-party inference framework for large transformers

Ke Cheng, Yuheng Xia, Anxiao Song, Jiaxuan Fu, Wenjie Qu, Yulong Shen, and Jiaheng Zhang. Mosformer: Maliciously secure three-party inference framework for large transformers. InProceedings of the ACM Conference on Computer and Communications Security, pages 4124– 4138, 2025

work page 2025
[59]

Scalable zero-knowledge proofs for non-linear functions in machine learning

Meng Hao, Hanxiao Chen, Hongwei Li, Chenkai Weng, Yuan Zhang, Haomiao Yang, and Tianwei Zhang. Scalable zero-knowledge proofs for non-linear functions in machine learning. In USENIX Security Symposium, 2024

work page 2024
[60]

zkLLM: Zero knowledge proofs for large language models

Haochen Sun, Jason Li, and Hongyang Zhang. zkLLM: Zero knowledge proofs for large language models. InProceedings of the ACM Conference on Computer and Communications Security, pages 4405–4419, 2024

work page 2024
[61]

Mingshu Cong, Sherman S. M. Chow, Siu-Ming Yiu, and Tsz Hon Yuen. Scalable zkSNARKs for matrix computations - A generic framework for verifiable deep learning. InProceedings of the International Conference on the Theory and Application of Cryptology and Information Security, Part V, pages 363–395, 2025

work page 2025
[62]

Chi, Quoc V

Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Brian Ichter, Fei Xia, Ed H. Chi, Quoc V. Le, and Denny Zhou. Chain-of-thought prompting elicits reasoning in large language models. InProceedings of the 36th International Conference on Neural Information Processing Systems, NIPS ’22, Red Hook, NY, USA, 2022. Curran Associates Inc. ISBN 9781713871088

work page 2022
[63]

Toolformer: Language models can teach themselves to use tools.Advances in Neural Information Processing Systems, 36:68539– 68551, 2023

Timo Schick, Jane Dwivedi-Yu, Roberto Dessì, Roberta Raileanu, 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

work page 2023
[64]

Gorilla: Large language model connected with massive apis.Advances in Neural Information Processing Systems, 37: 126544–126565, 2024

Shishir G Patil, Tianjun Zhang, Xin Wang, and Joseph E Gonzalez. Gorilla: Large language model connected with massive apis.Advances in Neural Information Processing Systems, 37: 126544–126565, 2024

work page 2024
[65]

Hugginggpt: Solving ai tasks with chatgpt and its friends in hugging face.Advances in Neural Information Processing Systems, 36:38154–38180, 2023

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

work page 2023
[66]

ToolLLM: Facilitating large language models to master 16000+ real-world APIs

Yujia Qin, Shihao Liang, Yining Ye, Kunlun Zhu, Lan Yan, Yaxi Lu, Yankai Lin, Xin Cong, Xiangru Tang, Bill Qian, et al. ToolLLM: Facilitating large language models to master 16000+ real-world APIs. InThe Twelfth International Conference on Learning Representations, 2024. 35 Holos

work page 2024
[67]

MemGPT: Towards LLMs as operating systems

Charles Packer, Vivian Fang, Shishir_G Patil, Kevin Lin, Sarah Wooders, and Joseph_E Gon- zalez. MemGPT: Towards LLMs as operating systems. 2023

work page 2023
[68]

Re- flexion: Language agents with verbal reinforcement learning.Advances in Neural Information Processing Systems, 36:8634–8652, 2023

Noah Shinn, Federico Cassano, Ashwin Gopinath, Karthik Narasimhan, and Shunyu Yao. Re- flexion: Language agents with verbal reinforcement learning.Advances in Neural Information Processing Systems, 36:8634–8652, 2023

work page 2023
[69]

Voyager: An open-ended embodied agent with large language models

Guanzhi Wang, Yuqi Xie, Yunfan Jiang, Ajay Mandlekar, Chaowei Xiao, Yuke Zhu, Linxi Fan, and Anima Anandkumar. Voyager: An open-ended embodied agent with large language models. Transactions on Machine Learning Research, 2024

work page 2024
[70]

Rolellm: Benchmarking, eliciting, and enhancing role-playing abilities of large language models

Noah Wang, Zy Peng, Haoran Que, Jiaheng Liu, Wangchunshu Zhou, Yuhan Wu, Hongcheng Guo, Ruitong Gan, Zehao Ni, Jian Yang, et al. Rolellm: Benchmarking, eliciting, and enhancing role-playing abilities of large language models. InFindings of the Association for Computational Linguistics: ACL 2024, pages 14743–14777, 2024

work page 2024
[71]

Camel: Communicative agents for" mind" exploration of large language model society.Advances in Neural Information Processing Systems, 36:51991–52008, 2023

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

work page 2023
[72]

Toward a Safe Internet of Agents

Juan A Wibowo and George C Polyzos. Toward a safe internet of agents.arXiv preprint arXiv:2512.00520, 2025

work page internal anchor Pith review Pith/arXiv arXiv 2025
[73]

AgentVerse: Facilitating multi-agent collaboration and exploring emergent behaviors, 2023

Weize Chen, Yusheng Su, Jingwei Zuo, Cheng Yang, Chenfei Yuan, Chi-Min Chan, Heyang Yu, Yaxi Lu, Yi-Hsin Hung, Chen Qian, Yujia Qin, Xin Cong, Ruobing Xie, Zhiyuan Liu, Maosong Sun, and Jie Zhou. AgentVerse: Facilitating multi-agent collaboration and exploring emergent behaviors, 2023

work page 2023
[74]

DyFlow: Dynamic workflow framework for agentic reasoning, 2025

Yanbo Wang, Zixiang Xu, Yue Huang, Xiangqi Wang, Zirui Song, Lang Gao, Chenxi Wang, Xiangru Tang, Yue Zhao, Arman Cohan, Xiangliang Zhang, and Xiuying Chen. DyFlow: Dynamic workflow framework for agentic reasoning, 2025

work page 2025
[75]

Multi- agent collaboration via evolving orchestration, 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, 2025

work page 2025
[76]

AgentNet: Decentralized evolutionary coordination for LLM-based multi-agent systems, 2025

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, 2025

work page 2025
[77]

Assemble your crew: Automatic multi-agent communication topology design via autoregressive graph generation, 2025

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

work page 2025
[78]

Latent collabo- ration in multi-agent systems, 2025

Jiaru Zou, Xiyuan Yang, Ruizhong Qiu, Gaotang Li, Katherine Tieu, Pan Lu, Ke Shen, Hang- hang Tong, Yejin Choi, Jingrui He, James Zou, Mengdi Wang, and Ling Yang. Latent collabo- ration in multi-agent systems, 2025

work page 2025
[79]

MasRouter: Learning to route LLMs for multi-agent systems, 2025

Yanwei Yue, Guibin Zhang, Boyang Liu, Guancheng Wan, Kun Wang, Dawei Cheng, and Yiyan Qi. MasRouter: Learning to route LLMs for multi-agent systems, 2025

work page 2025
[80]

AgentRouter: A knowledge-graph-guided LLM router for collaborative multi-agent question answering, 2025

Zheyuan Zhang, Kaiwen Shi, Zhengqing Yuan, Zehong Wang, Tianyi Ma, Keerthiram Muruge- san, Vincent Galassi, Chuxu Zhang, and Yanfang Ye. AgentRouter: A knowledge-graph-guided LLM router for collaborative multi-agent question answering, 2025

work page 2025

Showing first 80 references.