pith. machine review for the scientific record. sign in

arxiv: 2604.04753 · v1 · submitted 2026-04-06 · 📡 eess.SY · cs.SY

Recognition: no theorem link

Toward Self-Organizing Production Logistics in Circular Factories: A Multi-Agent Approach

Jan-Felix Klein , Yongkuk Jeong , Erik Flores-Garc\'ia , Magnus Wiktorsson
Authors on Pith no claims yet

Pith reviewed 2026-05-10 19:24 UTC · model grok-4.3

classification 📡 eess.SY cs.SY
keywords multi-agent systemsself-organizing logisticscircular factoriesproduction logisticsdecentralized decision-makingdigital twinsstructural uncertainty
0
0 comments X

The pith

A multi-agent system lets embodied agents negotiate tasks and responsibilities to create self-organizing production logistics in circular factories.

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

The paper proposes a vision for decentralized multi-agent coordination in production logistics where agents interpret tasks, assess their own capabilities, and negotiate assignments through event-driven communication. This is intended to replace rigid central planning in circular factories that face unpredictable product quality, availability, and timing. A reader would care because such variability makes traditional deterministic control unreliable, and the approach is claimed to deliver greater responsiveness and resilience by moving decisions closer to execution. The architecture combines embodied agents with a shared semantic knowledge layer and on-demand digital twins for monitoring and what-if evaluation. A three-phase roadmap is sketched using the self-organizing logistics typology to move from current practice toward full self-organization.

Core claim

The paper envisions a multi-agent system architecture that integrates embodied agents, a shared semantic knowledge layer, and dynamically instantiated digital twins to support decentralized decision-making through negotiations and event-driven communication, thereby enabling self-organizing production logistics in circular factories.

What carries the argument

Embodied agents that interpret tasks, assess capabilities, and negotiate responsibilities, supported by event-driven communication and a shared semantic knowledge layer.

If this is right

  • Decision-making moved closer to execution increases responsiveness to disruptions in product-core quality and availability.
  • Negotiation among agents improves resilience by allowing dynamic reallocation of responsibilities without central replanning.
  • The three-phase roadmap supplies a concrete sequence for progressing from conventional systems to full self-organizing production logistics.
  • Integration of digital twins enables real-time prediction and scenario evaluation to inform agent negotiations.

Where Pith is reading between the lines

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

  • Factories adopting this style of coordination may require less detailed advance scheduling than conventional systems.
  • The same negotiation mechanism could be tested in adjacent domains such as adaptive supply-chain rerouting under uncertainty.
  • Empirical validation would need metrics for coordination success rate and recovery time after disruptions.

Load-bearing premise

Decentralized negotiations and event-driven communication among agents can coordinate complex logistics and manage uncertainty without central oversight or causing coordination breakdowns.

What would settle it

A controlled pilot deployment in which agents repeatedly fail to reach timely agreements during a simulated quality or timing disruption, resulting in measurable production delays.

Figures

Figures reproduced from arXiv: 2604.04753 by Erik Flores-Garc\'ia, Jan-Felix Klein, Magnus Wiktorsson, Yongkuk Jeong.

Figure 1
Figure 1. Figure 1: Methodology linking emerging drivers, the system vision, and a [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Recent industrial pilot deployments of humanoid robots performing pro [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Conceptual vision of an SOPL system in the circular factory, combining [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: A three-phase development roadmap toward self-organizing production [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Phase 1 experimental setup in the IPU Lab at KTH, where heterogeneous [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
read the original abstract

Production logistics in circular factories is characterized by structural uncertainty due to variability in product-core quality, availability, and timing. These conditions challenge conventional deterministic and centrally planned control approaches. This paper proposes a vision for a multi-agent system based on decentralized decision-making through negotiations and event-driven communication serving as an enabler for self-organizing production logistics (SOPL) in circular factories. The envisioned system architecture integrates embodied agents, a shared semantic knowledge layer, and dynamically instantiated digital twins to support monitoring, prediction, and scenario evaluation. By shifting decision-making closer to execution and enabling agents to interpret tasks, assess capabilities, and negotiate responsibilities, the approach is expected to increase responsiveness and improve resilience to disruptions inherent in circular factories. Building on this vision, a three-phase development roadmap is introduced and characterized using the self-organizing logistics (SOL) typology, providing a structured pathway toward the realization of SOPL in circular factories.

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

3 major / 2 minor

Summary. The paper presents a conceptual vision for self-organizing production logistics (SOPL) in circular factories, which face structural uncertainty from variable product-core quality, availability, and timing. It proposes a multi-agent architecture using decentralized negotiations, event-driven communication, embodied agents, a shared semantic knowledge layer, and dynamically instantiated digital twins to enable monitoring, prediction, and scenario evaluation. Decision-making is shifted closer to execution so agents can interpret tasks, assess capabilities, and negotiate responsibilities, with expected gains in responsiveness and resilience. A three-phase development roadmap is outlined and mapped onto the self-organizing logistics (SOL) typology.

Significance. If the proposed architecture can be realized with provable coordination properties, it would address a genuine gap between conventional central planning and the needs of circular manufacturing systems. The integration of embodied agents with semantic layers and digital twins is a plausible direction, and the SOL-typology roadmap provides a structured way to track progress. However, the manuscript currently offers only high-level expectations rather than any demonstrated mechanism or evidence.

major comments (3)
  1. [§3] §3 (Architecture): The negotiation protocol among embodied agents is described only at the level of 'interpret tasks, assess capabilities, and negotiate responsibilities.' No formal specification of the protocol, message types, utility functions, or termination conditions is given, so it is impossible to assess whether negotiations are guaranteed to converge to stable allocations under product variability.
  2. [§3 and §4] §3 (Architecture) and §4 (Roadmap): The central claim that the approach 'is expected to increase responsiveness and improve resilience' is stated without any supporting model, simulation, or analytical bound. No analysis of coordination failure modes (negotiation deadlocks, inconsistent capability assessments, or cascading disruptions) appears, leaving the weakest assumption unexamined.
  3. [§4] §4 (Roadmap): The three-phase plan is characterized only by high-level SOL typology labels. No concrete milestones, performance metrics, or validation criteria are defined for Phase 1 or Phase 2, making it difficult to judge whether the roadmap can actually move the vision from conceptual to operational.
minor comments (2)
  1. [Abstract and Introduction] The abstract and introduction repeat the same high-level benefits without distinguishing what is novel relative to existing multi-agent manufacturing literature.
  2. [Figures and §4] Figure captions and the SOL typology table would benefit from explicit cross-references to the three phases so readers can map the roadmap directly onto the typology.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed comments, which help clarify the scope and limitations of our vision paper. We address each major comment below, noting that the manuscript proposes a conceptual architecture and roadmap rather than a fully implemented or validated system. Revisions will be made to improve clarity without altering the paper's visionary nature.

read point-by-point responses

  1. Referee: [§3] §3 (Architecture): The negotiation protocol among embodied agents is described only at the level of 'interpret tasks, assess capabilities, and negotiate responsibilities.' No formal specification of the protocol, message types, utility functions, or termination conditions is given, so it is impossible to assess whether negotiations are guaranteed to converge to stable allocations under product variability.

    Authors: The manuscript presents a high-level conceptual vision for the architecture rather than a detailed protocol specification. The phrasing in §3 is intentionally abstract to outline the intended agent behaviors and their integration with semantic knowledge and digital twins. Formal elements such as message types, utility functions, and convergence guarantees are beyond the current scope and are positioned for development in Phase 2 of the roadmap. We will revise §3 to explicitly state the conceptual level of the description and highlight the specific aspects (e.g., negotiation termination under variability) that require formalization in future work. revision: partial

  2. Referee: [§3 and §4] §3 (Architecture) and §4 (Roadmap): The central claim that the approach 'is expected to increase responsiveness and improve resilience' is stated without any supporting model, simulation, or analytical bound. No analysis of coordination failure modes (negotiation deadlocks, inconsistent capability assessments, or cascading disruptions) appears, leaving the weakest assumption unexamined.

    Authors: We acknowledge that the expectations of increased responsiveness and resilience are not backed by models, simulations, or failure-mode analysis in the manuscript. As this is a vision paper, these statements reflect design rationale drawn from decentralized multi-agent principles and related self-organizing systems literature, rather than empirical evidence. We will revise the relevant passages in §3 and §4 to frame the benefits as hypotheses to be tested, and we will add a brief discussion of potential coordination challenges such as deadlocks and inconsistent assessments as topics for investigation in the roadmap phases. revision: yes

  3. Referee: [§4] §4 (Roadmap): The three-phase plan is characterized only by high-level SOL typology labels. No concrete milestones, performance metrics, or validation criteria are defined for Phase 1 or Phase 2, making it difficult to judge whether the roadmap can actually move the vision from conceptual to operational.

    Authors: The roadmap in §4 is structured using the SOL typology to align with established self-organizing logistics research and to indicate progression from conceptual to operational stages. We agree that the current description lacks sufficient concreteness. We will expand §4 to include example milestones (e.g., semantic layer prototype for Phase 1, negotiation protocol simulation for Phase 2), suggested performance metrics (such as task allocation latency and resilience under core variability), and high-level validation criteria for each phase. revision: yes

Circularity Check

0 steps flagged

No circularity: vision proposal without derivations or self-referential reductions

full rationale

The paper is a conceptual vision document that outlines a high-level multi-agent architecture, semantic layer, digital twins, and a three-phase roadmap for self-organizing production logistics. It contains no equations, no formal models, no parameter fitting, and no claimed predictions that reduce to inputs by construction. The central expectations about responsiveness and resilience are presented as forward-looking outcomes of the proposed decentralized approach rather than derived results. No self-citation is used to justify a uniqueness theorem or to smuggle in an ansatz that bears the load of the argument. The SOL typology is referenced only to characterize the roadmap, without any reduction of the paper's claims to that typology by definition. The work is therefore self-contained as a proposal with no detectable circular steps.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 3 invented entities

The proposal rests on assumptions about the feasibility of agent negotiations and the utility of digital twins in this context, without providing new evidence for these.

axioms (2)
  • domain assumption Decentralized negotiations among agents can lead to effective coordination in uncertain environments
    Assumed in the proposal for self-organizing behavior.
  • domain assumption Digital twins can accurately support prediction and scenario evaluation in variable product conditions
    Central to the architecture for monitoring and evaluation.
invented entities (3)
  • Embodied agents no independent evidence
    purpose: Represent physical entities in the factory for decision-making
    Introduced as part of the system architecture.
  • Shared semantic knowledge layer no independent evidence
    purpose: Enable common understanding among agents
    Proposed component for interoperability.
  • Dynamically instantiated digital twins no independent evidence
    purpose: Support monitoring, prediction, and scenario evaluation
    Key element for handling uncertainty.

pith-pipeline@v0.9.0 · 5470 in / 1577 out tokens · 68165 ms · 2026-05-10T19:24:29.762998+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

33 extracted references · 23 canonical work pages · 1 internal anchor

  1. [1]

    Local path opti- mization in the latent space using learned distance gradient

    AgiBot World Colosseo: A Large-Scale Manipulation Platform for Scalable and Intelligent Embodied Systems. In: 2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). pp. 3549–3556. IEEE (2025). https://doi. org/10.1109/IROS60139.2025.11247088

    work page doi:10.1109/iros60139.2025.11247088 2025

  2. [2]

    https: //www.agilityrobotics.com/content/digit-moves-over-100k-totes (2025), accessed: 2026-03-15

    Agility Robotics: Digit moves over 100,000 totes in commercial deployment. https: //www.agilityrobotics.com/content/digit-moves-over-100k-totes (2025), accessed: 2026-03-15

    2025

  3. [3]

    https: //www.bmwgroup.com/en/news/general/2026/humanoid-robot-in-leipzig.html (2026), accessed: 2026-03-15

    BMW Group: First humanoid robot introduced in plant leipzig. https: //www.bmwgroup.com/en/news/general/2026/humanoid-robot-in-leipzig.html (2026), accessed: 2026-03-15

    2026

  4. [4]

    (ed.): Self-organization in biological systems

    Camazine, S. (ed.): Self-organization in biological systems. Princeton studies in complexity, Princeton University Press, Princeton, NJ, 2nd print., and 1st paper- back print edn. (2003)

    2003

  5. [5]

    Journal of Manufacturing Systems 59, 386–397 (2021)

    Diaz C., J.L., Ocampo-Martinez, C.: Non-centralised control strategies for energy- efficient and flexible manufacturing systems. Journal of Manufacturing Systems 59, 386–397 (2021). https://doi.org/10.1016/j.jmsy.2021.02.004 14 J.-F. Klein et al

    work page doi:10.1016/j.jmsy.2021.02.004 2021

  6. [6]

    https://environment.ec

    European Commission: Circular economy – environment. https://environment.ec. europa.eu/strategy/circular-economy_en (2025), accessed 2026-02-10 and Circular Economy Act due for adoption in 2026

    2025

  7. [7]

    https:// www.figure.ai/news/production-at-bmw (2025), accessed: 2026-03-14

    Figure AI: F.02 contributed to the production of 30000 cars at BMW. https:// www.figure.ai/news/production-at-bmw (2025), accessed: 2026-03-14

    2025

  8. [8]

    at - Au- tomatisierungstechnik72(9), 861–874 (2024)

    Fleischer, J., Zanger, F., Schulze, V., Neumann, G., Stricker, N., Furmans, K., Pfrommer, J., Lanza, G., Hansjosten, M., Fischmann, P., Dvorak, J., Klein, J.F., Rauscher, F., Ebner, A., May, M.C., Gönnheimer, P.: Self-learning and au- tonomously adapting manufacturing equipment for the circular factory. at - Au- tomatisierungstechnik72(9), 861–874 (2024)....

    work page doi:10.1515/auto-2024- 2024

  9. [9]

    PhD, University of Twente, Enschede, The Netherlands (2023)

    Gerrits, B.: Self-Organizing Logistics : Towards a unifying framework for auto- mated transport systems. PhD, University of Twente, Enschede, The Netherlands (2023). https://doi.org/10.3990/1.9789036555555

    work page doi:10.3990/1.9789036555555 2023

  10. [10]

    International Transactions in Opera- tional Research31(3), 1309–1374 (2024)

    Gerrits, B., Van Heeswijk, W., Mes, M.: Towards self-organizing logistics in trans- portation: a literature review and typology. International Transactions in Opera- tional Research31(3), 1309–1374 (2024). https://doi.org/10.1111/itor.13408

    work page doi:10.1111/itor.13408 2024

  11. [11]

    at - Automatisierung- stechnik72(9), 875–883 (2024)

    Hofmann, C., Staab, S., Selzer, M., Neumann, G., Furmans, K., Heizmann, M., Beyerer, J., Lanza, G., Pfrommer, J., Düser, T., Klein, J.F.: The role of an ontology-based knowledge backbone in a circular factory. at - Automatisierung- stechnik72(9), 875–883 (2024). https://doi.org/10.1515/auto-2024-0006, https: //www.degruyter.com/document/doi/10.1515/auto-2...

    work page doi:10.1515/auto-2024-0006 2024

  12. [12]

    https://thehumanoid.ai/humanoid-and-siemens- completed-a-proof-of-concept-to-test-humanoidrobots-in-industrial-logistics/ (2026), accessed: 2026-03-14

    Humanoid: Humanoid and siemens completed a proof of concept to test humanoid robots in industrial logistics. https://thehumanoid.ai/humanoid-and-siemens- completed-a-proof-of-concept-to-test-humanoidrobots-in-industrial-logistics/ (2026), accessed: 2026-03-14

    2026

  13. [13]

    International Federation of Robotics: World robotics report 2025: Service robots (2025), https://ifr.org/worldrobotics, accessed: 2026-01-28

    2025

  14. [14]

    American Institute of Physics Conference Series, vol

    Ismayyir, D.K., Dawood, L.M., AL-Khafaji, M.M.H.: Modelling and control archi- tecturesofproductionsystems:Literaturereview.In:AmericanInstituteofPhysics Conference Series. American Institute of Physics Conference Series, vol. 3079, p. 060022. AIP (2024). https://doi.org/10.1063/5.0202238

    work page doi:10.1063/5.0202238 2024

  15. [15]

    https://bostondynamics.com/blog/ getting-real-with-humanoids/ (2025), accessed: 2026-03-15

    Jackowski, Z.: Getting real with humanoids. https://bostondynamics.com/blog/ getting-real-with-humanoids/ (2025), accessed: 2026-03-15

    2025

  16. [16]

    Procedia CIRP120, 368–373 (2023)

    Klein, J.F., Furmans, K.: Digital twin architecture and sim-to-real gap analysis of a material transfer system in a remanufacturing environment. Procedia CIRP120, 368–373 (2023). https://doi.org/10.1016/j.procir.2023.09.004

    work page doi:10.1016/j.procir.2023.09.004 2023

  17. [17]

    Logistics Journal: Proceedings (21) (2025)

    Klein, J.F., Wolf, R., Ernst, A., Shi, Y., Schumacher, P., Thapa, R.B., Rayyes, R., Furmans, K.: A Knowledge-Based Intralogistic System for a Circular Factory. Logistics Journal: Proceedings (21) (2025). https://doi.org/10.2195/LJ_PROC_ KLEIN_202510_01

    work page doi:10.2195/lj_proc_ 2025

  18. [18]

    at - Automatisierung- stechnik70(6), 504–516 (2022)

    Lanza, G., Asfour, T., Beyerer, J., Deml, B., Fleischer, J., Heizmann, M., Fur- mans, K., Hofmann, C., Cebulla, A., Dreher, C., Kaiser, J.P., Klein, J.F., Leven, F., Mangold, S., Mitschke, N., Stricker, N., Pfrommer, J., Wu, C., Wurster, M., Zaremski, M.: Agiles Produktionssystem mittels lernender Roboter bei ungewissen Produktzuständen am Beispiel der An...

    work page doi:10.1515/auto-2021-0158 2022

  19. [19]

    at - Automa- tisierungstechnik72(9), 774–788 (2024)

    Lanza, G., Deml, B., Matthiesen, S., Martin, M., Brützel, O., Hörsting, R.: The vision of the circular factory for the perpetual innovative product. at - Automa- tisierungstechnik72(9), 774–788 (2024). https://doi.org/10.1515/auto-2024-0012 Toward Self-Organizing Production Logistics 15

    work page doi:10.1515/auto-2024-0012 2024

  20. [20]

    Vicinagearth1(9) (2024).https://doi

    Li, X., Wang, S., Zeng, S., Wu, Y., Yang, Y.: A survey on LLM-based multi-agent systems: workflow, infrastructure, and challenges. Vicinagearth1(1), 9 (2024). https://doi.org/10.1007/s44336-024-00009-2

    work page doi:10.1007/s44336-024-00009-2 2024

  21. [21]

    LogisticsIQ: AGV-AMR Market (2025), https://www.thelogisticsiq.com/research/ automated-guided-vehicles-agv-market/, accessed: 2026-01-28

    2025

  22. [22]

    Fajemisin AO, Maragno D, den Hertog D

    Mes,M.,VanDerHeijden,M.,VanHarten,A.:Comparisonofagent-basedschedul- ing to look-ahead heuristics for real-time transportation problems. European Jour- nal of Operational Research181(1), 59–75 (2007). https://doi.org/10.1016/j.ejor. 2006.02.051

    work page doi:10.1016/j.ejor 2007

  23. [23]

    IEEE Transactions on Knowledge and Data Engineering36(7), 3580–3599 (2024)

    Pan, S., Luo, L., Wang, Y., Chen, C., Wang, J., Wu, X.: Unifying Large Language Models and Knowledge Graphs: A Roadmap. IEEE Transactions on Knowledge and Data Engineering36(7), 3580–3599 (2024). https://doi.org/10.1109/TKDE. 2024.3352100

    work page doi:10.1109/tkde 2024

  24. [24]

    CRC Press, 1 edn

    Pariès, J., Wreathall, J.: Resilience Engineering in Practice: A Guidebook. CRC Press, 1 edn. (May 2017). https://doi.org/10.1201/9781317065265

    work page doi:10.1201/9781317065265 2017

  25. [25]

    at - Automatisierungstechnik70(6), 534–541 (2022)

    Pfrommer, J., Klein, J.F., Wurster, M., Rapp, S., Grauberger, P., Lanza, G., Al- bers, A., Matthiesen, S., Beyerer, J.: An ontology for remanufacturing systems. at - Automatisierungstechnik70(6), 534–541 (2022). https://doi.org/10.1515/auto- 2021-0156

    work page doi:10.1515/auto- 2022

  26. [26]

    The International Journal of Advanced Manufacturing Technology84(1-4), 147– 164 (2016)

    Qu, T., Lei, S.P., Wang, Z.Z., Nie, D.X., Chen, X., Huang, G.Q.: IoT-based real- timeproductionlogisticssynchronizationsystemundersmartcloudmanufacturing. The International Journal of Advanced Manufacturing Technology84(1-4), 147– 164 (2016). https://doi.org/10.1007/s00170-015-7220-1

    work page doi:10.1007/s00170-015-7220-1 2016

  27. [27]

    Scientific Reports15(1), 13755 (2025)

    Raza, M., Jahangir, Z., Riaz, M.B., Saeed, M.J., Sattar, M.A.: Industrial appli- cations of large language models. Scientific Reports15(1), 13755 (2025). https: //doi.org/10.1038/s41598-025-98483-1

    work page doi:10.1038/s41598-025-98483-1 2025

  28. [28]

    Ad- vances in Manufacturing5(4), 344–358 (2017)

    Strandhagen, J.W., Alfnes, E., Strandhagen, J.O., Vallandingham, L.R.: The fit of Industry 4.0 applications in manufacturing logistics: a multiple case study. Ad- vances in Manufacturing5(4), 344–358 (2017). https://doi.org/10.1007/s40436- 017-0200-y

    work page doi:10.1007/s40436- 2017

  29. [29]

    Multi-Agent Collaboration Mechanisms: A Survey of LLMs

    Tran, K.T., Dao, D., Nguyen, M.D., Pham, Q.V., O’Sullivan, B., Nguyen, H.D.: Multi-Agent Collaboration Mechanisms: A Survey of LLMs (2025). https://doi. org/10.48550/ARXIV.2501.06322

    work page internal anchor Pith review doi:10.48550/arxiv.2501.06322 2025

  30. [30]

    UBTECH Robotics: Unleashing swarm intelligence: UBTECH pioneers the world’s first multi-humanoid robot collaborative training in multi-task, multi-scenario set- tings at ZEEKR. https://www.prnewswire.com/news-releases/unleashing-swarm- intelligence-ubtech-pioneers-the-worlds-first-multi-humanoid-robot-collaborative- training-in-multi-task-multi-scenario-s...

    2025

  31. [31]

    Sustainable Production and Consumption33, 1043–1058 (2022)

    Viles, E., Kalemkerian, F., Garza-Reyes, J.A., Antony, J., Santos, J.: Theoriz- ing the Principles of Sustainable Production in the context of Circular Economy and Industry 4.0. Sustainable Production and Consumption33, 1043–1058 (2022). https://doi.org/10.1016/j.spc.2022.08.024

    work page doi:10.1016/j.spc.2022.08.024 2022

  32. [32]

    In: Hülsmann, M., Windt, K

    Windt, K., Hülsmann, M.: Changing Paradigms in Logistics — Understanding the Shift from Conventional Control to Autonomous Cooperation and Control. In: Hülsmann, M., Windt, K. (eds.) Understanding Autonomous Cooperation and Control in Logistics, pp. 1–16. Springer Berlin Heidelberg, Berlin, Heidelberg (2007). https://doi.org/10.1007/978-3-540-47450-0_1

    work page doi:10.1007/978-3-540-47450-0_1 2007

  33. [33]

    https://doi.org/10.48550/ARXIV.2502.13130

    Yang, J., Tan, R., Wu, Q., Zheng, R., Peng, B., Liang, Y., Gu, Y., Cai, M., Ye, S., Jang, J., Deng, Y., Liden, L., Gao, J.: Magma: A Foundation Model for Multimodal AI Agents (2025). https://doi.org/10.48550/ARXIV.2502.13130

    work page doi:10.48550/arxiv.2502.13130 2025