Pith. sign in

REVIEW 3 major objections 4 minor 6 cited by

This paper proposes Agentic Services Computing: a lifecycle- and capability-driven paradigm that treats intelligent LLM-based agents as first-class, governable services rather than passive endpoints.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-04 14:32 UTC pith:MTBNUKTD

load-bearing objection A well-organized survey that makes a plausible case for a new subfield, but the core governance mechanism is a promise, not a result. the 3 major comments →

arxiv 2509.24380 v3 pith:MTBNUKTD submitted 2025-09-29 cs.SE

Agentic Services Computing

Shuiguang Deng , Hailiang Zhao , Ziqi Wang , Wenzhuo Qian , Xiang Ao , Guanjie Cheng , Jianwei Yin , Albert Y. Zomaya
show 1 more author
Schahram Dustdar
This is my paper
classification cs.SE
keywords agentic services computingLLM agentsmulti-agent systemsservice lifecycleservice governancetrustworthinessautonomous decision-makingservice-oriented architecture
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper argues that large language model-powered agents represent a new kind of service—autonomous, goal-driven, stateful, and collaborative—that existing service paradigms (SOA, microservices, serverless) and existing agent research each handle only partially. It introduces Agentic Services Computing (ASC) as a unifying paradigm, formalized in Definition 1, that brings agents into the full service lifecycle: design, deployment, operation, and evolution. The work maps four research dimensions—perception, decision-making, multi-agent collaboration, and evaluation/alignment/trustworthiness—onto that lifecycle, producing a research agenda rather than a mature implementation. A sympathetic reader would care because, if ASC is right, the engineering discipline that made Web services reliable and governable can be extended to autonomous agents, making them auditable, contract-bound, and evolvable in production systems.

Core claim

The central claim is Definition 1: Agentic Services Computing is a lifecycle- and capability-driven paradigm for engineering intelligent agents as first-class services. The paper asserts that agentic services—persistent, goal-oriented agents that perceive, reason, act, and evolve—can be described, composed, delivered, monitored, audited, and retired using structures inherited from services computing, now extended with cognitive autonomy and emergent collaboration. It organizes the field along two axes: the four-phase lifecycle (Design, Deployment, Operation, Evolution) and four capability pillars (perception and context modeling; autonomous decision-making; multi-agent collaboration; evaluat

What carries the argument

The organizing mechanism is the ASC framework matrix: a lifecycle axis (Design, Deployment, Operation, Evolution) crossed with four capability dimensions (Perception/Context; Decision-Making/Execution; Multi-Agent Collaboration; Evaluation/Alignment/Trustworthiness). The formal anchor is Definition 1, which defines ASC as a 'lifecycle- and capability-driven paradigm for engineering intelligent agents as first-class services.' Supporting the definition are the SCALE characteristics, which specify what distinguishes an agentic service from a traditional endpoint. This matrix does the argumentative work: it converts scattered agent techniques into a structured service-engineering space, so that

Load-bearing premise

The paradigm stands or falls on whether LLM-based agents can be made reliable, interpretable, and deterministic enough that service-level contracts, audit trails, versioning, and safe retirement genuinely apply to them in practice.

What would settle it

A concrete test: deploy a set of production agentic services under ASC-style ASLOs (alignment fidelity, interpretability thresholds, bounded actions) for a quarter and measure the fraction of tasks where the reasoning-trace audit fails to explain an observed failure or policy violation; if that fraction stays high, the audit/governance premise fails. Alternatively, an experiment where versioned rollback of an agent's prompt or policy does not restore prior behavior would directly undercut the lifecycle-governance claim.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Agents can be brought under the same lifecycle governance as traditional services, including versioned prompts, policies, memory schemas, canary deployments, and safe retirement protocols.
  • Agent Service Level Objectives (ASLOs) become machine-enforceable contracts covering not only latency and availability but also alignment fidelity, safety boundaries, and interpretability.
  • Monitoring extends from system metrics to cognitive observability: logging reasoning traces and decision pathways, not just API calls.
  • Multi-agent collaboration can be viewed as dynamic service composition, with negotiation, reputation, and fault tolerance replacing static orchestration.
  • Trustworthiness is treated as a lifelong, cross-cutting property enforced through audits, constitutional constraints, and human-in-the-loop governance, not a post-hoc add-on.

Where Pith is reading between the lines

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

  • If ASC is adopted, the unit of service contracting shifts from 'what function to call' to 'what goal to achieve,' which may require new liability and compliance frameworks for outcomes rather than outputs.
  • The same lifecycle machinery could be applied to non-LLM agents (e.g., robotic or classical BDI agents), giving ASC a broader scope than the LLM-centric examples suggest, though the paper does not develop this.
  • The paper's roadmap implies a testable intermediate milestone: before interoperable agent societies arrive, one could validate ASLOs on a small set of production agents by measuring whether audit-trail completeness correlates with failure recovery speed.
  • The framework suggests that agent registries and audit APIs could become standardized infrastructure akin to package registries or service meshes; a concrete extension would be a reference implementation of a registry with versioned prompts and reasoning-trace hashes.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper proposes Agentic Services Computing (ASC), a paradigm and research framework for treating LLM-based agents as first-class services. It surveys the foundations in services computing, multi-agent systems, and LLM-based agents; introduces the SCALE characteristics of agentic services; proposes a four-phase lifecycle (Design, Deployment, Operation, Evolution); and organizes research into four dimensions: perception/context modeling, autonomous decision-making, multi-agent collaboration, and evaluation/alignment/trustworthiness. It concludes with a three-horizon roadmap and future trends. The paper is a position/survey contribution rather than an empirical validation of the framework.

Significance. If the proposed framework holds, it could provide a valuable unifying synthesis for a fragmented literature, connecting classical MAS principles, LLM agent capabilities, and service-lifecycle engineering. The paper's strengths are its broad and current coverage, the lifecycle-by-capability matrix in Figure 4, the comparative protocol and benchmark tables, and its explicit identification of open challenges. The roadmap format and the honest acknowledgment of missing pieces are useful for the community. However, the contribution is conceptual and organizational, not evidential: there is no empirical validation, and the central governance claims depend on mechanisms that are described only as future milestones.

major comments (3)
  1. [§IV-C; Definition 1 (§IV-A)] The abstract and Definition 1 claim that ASC enables agentic services to be systematically described, composed, delivered, monitored, audited, and evolved as first-class services. This requires a verifiable service contract: an interface specification with observable predicates and satisfaction conditions. The paper does not provide one. §III-B mentions 'service contracts that align with SOA principles' but does not define what a contract is for an autonomous, non-deterministic agent. §IV-C defers ASLOs to a mid-term milestone, and §VIII-D concedes that post-deployment validation is lacking. This is a load-bearing gap, not merely an LLM-reliability concern. Please either add a concrete ASLO/contract sketch with machine-checkable predicates and an evaluation procedure, or explicitly reposition ASC as a research agenda whose operational core is a goal rather than an existing mechanism.
  2. [§VIII-C; §IV-C] The trustworthiness pillar relies on audit APIs that expose 'reasoning-action traces' and 'auditable decision logs.' However, the paper itself notes in Table VI that Rationale-Augmented Agents can produce plausible but factually incorrect justifications. If internal reasoning traces are not causally faithful, they do not provide the accountability that ASC's governance claims require. The paper should specify which observables are verifiable (e.g., tool invocations, state transitions, pre/post conditions, outputs) and which are not (e.g., natural-language rationales), and explain how ASLOs would be checked against the verifiable subset. Without an evidence model, the audit mechanism is asserted rather than defined.
  3. [§VIII-D; §IV-C] The paper's own summary states that evaluation, alignment, and trustworthiness are fragmented: evaluation often ignores alignment, alignment lacks post-deployment validation, and trust mechanisms are added post-hoc. Since the ASC lifecycle is explicitly closed-loop—operational feedback feeds redesign and evolution—this gap is structural rather than a routine list of future work. The milestones in §IV-C are not tied to validation criteria. Please connect each milestone to a concrete success criterion, for example an ASLO that is machine-checkable on production traces, so that the lifecycle claim can be tested or at least sharply scoped.
minor comments (4)
  1. [Table I] The 'Classical MAS' row cites reference [63], which is a survey on vision-based robotic grasping, not a survey of classical multi-agent systems. This citation error undermines the comparison table; please replace it with a proper MAS survey reference.
  2. [§IV-C] The sentence 'The following sections elaborate on each of the four research dimensions...' appears twice in consecutive paragraphs at the end of §IV-C. One occurrence should be deleted.
  3. [§VI-B.2] The text refers to 'APRMCTS [68]', but the cited reference title is 'Automated Repair of Programs from Large Language Models'. The label does not match the reference; please align the name or supply the intended citation.
  4. [§III-A] The SCALE framework is described as 'formalized,' but the five characteristics are qualitative. Consider using 'structured characterization' or another term that does not imply a formal specification, unless a formal definition is actually provided.

Circularity Check

0 steps flagged

No significant circularity: ASC is a definitional/organizational framework; its claims do not reduce to fitted inputs or self-citations.

full rationale

This paper is a survey/position paper: the central claim is Definition 1, which stipulates what Agentic Services Computing is, and the SCALE characteristics and four-phase lifecycle are presented as an organizing taxonomy rather than as consequences derived from prior equations or fitted data. The paper makes no quantitative predictions from fitted parameters, and no output is equivalent to an input by construction. The load-bearing external evidence consists of independently published agent systems, benchmarks, and standards (e.g., ReAct, Reflexion, AutoGen, MCP, A2A), which are used as examples and foundations, not as premises that presuppose ASC. The authors' self-citations (Refs. 57, 58, 199–202) appear in peripheral discussions of cloud-native background and scheduling examples; none is used to justify the ASC framework itself, to exclude alternatives, or to import a uniqueness theorem. Section IV-C explicitly defers ASLOs and agent contracts to mid-term milestones, and Section VIII-D admits evaluation/alignment integration remains fragmented; these are honest limitations, not disguised circular inputs. The framework's contribution is organizational and definitional, and no derivation loop was identified.

Axiom & Free-Parameter Ledger

0 free parameters · 3 axioms · 3 invented entities

The paper's central contribution is conceptual, not empirical, so the axiom ledger is dominated by domain assumptions about agent capability and governance feasibility. It introduces three invented entities that are terminological rather than falsifiable, and no free parameters are fit to data.

axioms (3)
  • domain assumption LLM-based agents can reliably perceive, reason, act, and collaborate in open environments.
    The entire premise of ASC is that agents are capable enough to be treated as autonomous services. This is argued through examples but not demonstrated in the paper (Sections II-C, VI).
  • domain assumption Service lifecycle governance (versioning, monitoring, SLAs, retirement) is applicable to autonomous agents.
    The paper assumes that governance mechanisms from services computing extend to non-deterministic agents. This is asserted in Sections II-D and IV-C, but no evidence is provided that such governance is feasible at scale.
  • domain assumption Classical MAS, services computing, and LLM-based agents are complementary and can be unified into one discipline.
    The synthesis claim underlying ASC, stated in Section II-D, requires that the three paradigms integrate without fundamental conflict. This is a reasonable but unproven position.
invented entities (3)
  • Agentic Service no independent evidence
    purpose: Defines the central unit of ASC, a persistent goal-driven agent treated as a service.
    Conceptual term introduced by the paper; no external falsifiable handle beyond the authors' definition.
  • SCALE framework no independent evidence
    purpose: Five characteristics (Stateful, Cognitive, Agent as goal executor, Lifelong collaborative learning, End-to-end lifecycle governance) that define an agentic service.
    An acronym/framework proposed by the authors to organize their definition.
  • Agent Service Level Objective (ASLO) no independent evidence
    purpose: Extension of SLA to specify alignment, safety, interpretability, and sustainability commitments for agents.
    Proposed metric in Section IV-C; not yet operationalized or measured anywhere.

pith-pipeline@v1.3.0-alltime-deepseek · 126 in / 7743 out tokens · 124020 ms · 2026-08-04T14:32:30.446110+00:00 · methodology

0 comments
read the original abstract

Services computing has evolved from Web services and microservices to cloud-native and serverless paradigms. These approaches established mature principles for describing, composing, deploying, operating, and governing reusable software functions. LLM-based agents now introduce a fundamentally different service form. Service value in this paradigm emerges not only from invoking predefined functions but also from delegating goals to autonomous entities. These entities understand context, use tools, collaborate with peers, and act across open environments. This shift raises a core question for services computing. How can goal-driven, stateful, tool-mediated, and accountable autonomous behavior be engineered and managed as a service? Recent studies on LLM agents and multi-agent systems provide important foundations. A clear service-centered research roadmap for this emerging paradigm nevertheless remains absent. This work introduces Agentic Services Computing (ASC) to address this gap. ASC extends services computing from managing reusable functional endpoints to engineering and governing autonomous service entities. It defines agentic services as service-oriented autonomous agents. Related research is organized through a lifecycle view that connects service objects, system structures, enabling infrastructure, evaluation metrics, application evidence, and open challenges. This service-centered perspective establishes a foundation for future service ecosystems. Autonomous agents can thus be systematically described, composed, delivered, monitored, audited, and evolved as first-class services.

Figures

Figures reproduced from arXiv: 2509.24380 by Albert Y. Zomaya, Guanjie Cheng, Hailiang Zhao, Jianwei Yin, Schahram Dustdar, Shuiguang Deng, Wenzhuo Qian, Xiang Ao, Ziqi Wang.

Figure 1
Figure 1. Figure 1: We define Agentic Services Computing as an emer [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The orchestration of core sections in this paper. [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Comparison between agentic and traditional services. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Research framework of Agentic Services Computing, mapping core research dimensions across the service lifecycle. Each [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Key dimensions of perception, context, and environ [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Core paradigms and mechanisms for the autonomous [PITH_FULL_IMAGE:figures/full_fig_p011_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Core paradigms and mechanisms for multi-agent col [PITH_FULL_IMAGE:figures/full_fig_p013_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: The Model layer captures shared domain seman￾Services / Frameworks Tools / Resources Without MCP With MCP MCP Protocol MCP Server A MCP Server N Data Remote Service MCP Protocol MCP Protocol … Host with MCP Client (Claude, IDEs) Web APIs MCP Architecture Server Host Client Communication Interface Data Exchange Database Connection API Access Task Execution Chat Application Code Assistant Architecture Compar… view at source ↗
Figure 9
Figure 9. Figure 9: Agent collaboration and task coordination via CNP. [PITH_FULL_IMAGE:figures/full_fig_p016_9.png] view at source ↗

discussion (0)

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

Forward citations

Cited by 6 Pith papers

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

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

    cs.MA 2026-05 unverdicted novelty 7.0

    DESBench reveals structural trade-offs among centralized, hierarchical, heterarchical, and holonic coordination in dynamic industrial scheduling that outcome metrics alone miss.

  2. Credibility Trilemma in Polymatroidal Service Markets

    cs.GT 2026-05 unverdicted novelty 5.0

    In non-modular polymatroidal service markets, revenue-optimal DSIC mechanisms cannot also be credible for strategic operators, with tight welfare-loss bounds on the Cost of Non-Credibility across network topologies.

  3. AgentOpt v0.1 Technical Report: Client-Side Optimization for LLM-Based Agent

    cs.LG 2026-04 unverdicted novelty 5.0

    AgentOpt introduces a framework-agnostic package that uses algorithms like UCB-E to find cost-effective model assignments in multi-step LLM agent pipelines, cutting evaluation budgets by 62-76% while maintaining near-...

  4. Industrial Data-Service-Knowledge Governance: Toward Integrated and Trusted Intelligence

    cs.CE 2026-01 conditional novelty 5.0

    TRISK is a taxonomy and architecture for governing industrial AI trust across data, service, and knowledge layers, backed by a survey of 120+ works and one unreported case study.

  5. Agentic Metaverse Services: A New As-a-Service Paradigm

    cs.SE 2026-07 conditional novelty 4.0

    Agent capabilities can be delivered as Meta-AaaS so that autonomous, embodied agents become the unit of service in metaverse business processes.

  6. OpenClaw and Ollama in Agentic AI: Toward Fully Autonomous and Scalable AI Agent Systems

    cs.AI 2026-04 reject novelty 3.0

    A review-plus-demo claiming agentic capabilities emerge from system integration, backed by a 15-task benchmark whose C1→C3 performance gap is largely built into the test design.

Reference graph

Works this paper leans on

199 extracted references · 26 linked inside Pith · cited by 6 Pith papers

  1. [1]

    https://airflow.apache.org/

    Apache airflow. https://airflow.apache.org/. Accessed: 2025-08-27

    2025

  2. [2]

    https://babyagi.org/

    Babyagi. https://babyagi.org/. Accessed: 2025-08-27

    2025

  3. [3]

    https://camunda.com/

    Camunda: The universal process orchestrator. https://camunda.com/. Accessed: 2025-09-03

    2025

  4. [4]

    https://grafana.com/

    Grafana: The open-source platform for monitoring and observability. https://grafana.com/. Accessed: 2025-08-27

    2025

  5. [5]

    https:// kubernetes.io/

    Kubernetes: Production-grade container orchestration. https:// kubernetes.io/. Accessed: 2025-08-27

    2025

  6. [6]

    https:// www.llamaindex.ai/

    Llamaindex: Redefine document workflows with ai agents. https:// www.llamaindex.ai/. Accessed: 2025-08-28

    2025

  7. [7]

    https://opentelemetry.io/

    Prometheus: High-quality, ubiquitous, and portable telemetry to enable effective observability. https://opentelemetry.io/. Accessed: 2025-08- 27

    2025

  8. [8]

    https://prometheus.io/

    Prometheus: Open source metrics and monitoring for your systems and services. https://prometheus.io/. Accessed: 2025-08-27

    2025

  9. [9]

    https://github.com/ microsoft/semantic-kernel

    Semantic kernel: Build intelligent ai agents and multi-agent systems with this enterprise-ready orchestration framework. https://github.com/ microsoft/semantic-kernel. Accessed: 2025-08-28

    2025

  10. [10]

    Autogpt: An autonomous gpt-4 experiment. 2023. GitHub Repository: https://github.com/DataBassGit/Auto-GPT. [11]The Agent2Agent Protocol, 2025. https://a2a-protocol.org/latest/

    2023

  11. [12]

    https://www

    Guardrails ai: Mitigate gen ai risks with guardrails, 2025. https://www. guardrailsai.com/. [13]HayStack: The Production-Ready Open Source AI Framework, 2025. https://haystack.deepset.ai/. [14]LangGraph: Multi-Actor Application Framework, 2025. https://www. langchain.com/langgraph

    2025

  12. [15]

    https://www.langchain

    Langsmith: Ship agents with confidence, 2025. https://www.langchain. com/langsmith. [16]Model Context Protocol, 2025. https://modelcontextprotocol.io/docs/ getting-started/intro. [17]OpenAI Function Calling Guide, 2025. https://platform.openai.com/ docs/guides/function-calling. [18]Orchestrate with Confidence: The Open-Source Workflow Engine for Tomorrow,...

    2025

  13. [19]

    https: //www.pinecone.io/

    Pinecone: The vector database for scale in production, 2025. https: //www.pinecone.io/. [20]TrustChain: Fostering a Human-Centred, Trustworthy and Sustainable Internet, 2025. https://trustchain.ngi.eu/

    2025

  14. [21]

    https:// weaviate.io/

    Weaviate: The ai-native database developers love, 2025. https:// weaviate.io/

    2025

  15. [22]

    Deep learning with differential privacy

    Martin Abadi, Andy Chu, Ian Goodfellow, H Brendan McMahan, Ilya Mironov, Kunal Talwar, and Li Zhang. Deep learning with differential privacy. InProceedings of the 2016 ACM SIGSAC conference on computer and communications security, pages 308–318, 2016

    2016

  16. [23]

    Gpt-4 technical report.arXiv preprint arXiv:2303.08774, 2023

    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

    Pith/arXiv arXiv 2023

  17. [24]

    Act-1: A generalist agent for tool use

    Adept AI. Act-1: A generalist agent for tool use. Technical report,

  18. [25]

    Flamingo: a visual language model for few- shot learning.Advances in neural information processing systems, 35:23716–23736, 2022

    Jean-Baptiste Alayrac, Jeff Donahue, Pauline Luc, Antoine Miech, Iain Barr, Yana Hasson, Karel Lenc, Arthur Mensch, Katherine Millican, Malcolm Reynolds, et al. Flamingo: a visual language model for few- shot learning.Advances in neural information processing systems, 35:23716–23736, 2022

    2022

  19. [26]

    Vision-and-language navigation: Interpreting visually-grounded navigation instructions in real environments

    Peter Anderson, Qi Wu, Damien Teney, Jake Bruce, Mark Johnson, Niko Sünderhauf, Ian Reid, Stephen Gould, and Anton Van Den Hen- gel. Vision-and-language navigation: Interpreting visually-grounded navigation instructions in real environments. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 3674–3683, 2018

    2018

  20. [27]

    https://argilla.io

    Argilla.Argilla, 2025. https://argilla.io

    2025

  21. [28]

    Constitutional ai: Harmlessness from ai feedback.arXiv preprint arXiv:2212.08073, 2022

    Yuntao Bai, Saurav Kadavath, Sandipan Kundu, Amanda Askell, Jack- son Kernion, Andy Jones, Anna Chen, Anna Goldie, Azalia Mirhoseini, Cameron McKinnon, et al. Constitutional ai: Harmlessness from ai feedback.arXiv preprint arXiv:2212.08073, 2022

    Pith/arXiv arXiv 2022

  22. [29]

    Serverless computing: Current trends and open problems

    Ioana Baldini, Paul Castro, Kerry Chang, Perry Cheng, Stephen Fink, Vatche Ishakian, Nick Mitchell, Vinod Muthusamy, Rodric Rabbah, Aleksander Slominski, et al. Serverless computing: Current trends and open problems. InResearch advances in cloud computing, pages 1–20. Springer, 2017

    2017

  23. [30]

    Beit: Bert pre- training of image transformers.arXiv preprint arXiv:2106.08254, 2021

    Hangbo Bao, Li Dong, Songhao Piao, and Furu Wei. Beit: Bert pre- training of image transformers.arXiv preprint arXiv:2106.08254, 2021

    Pith/arXiv arXiv 2021

  24. [31]

    Addison-Wesley Professional, 2015

    Len Bass, Ingo Weber, and Liming Zhu.DevOps: A software architect’s perspective. Addison-Wesley Professional, 2015

    2015

  25. [32]

    Jade–a fipa-compliant agent framework

    Fabio Bellifemine, Agostino Poggi, and Giovanni Rimassa. Jade–a fipa-compliant agent framework. InProceedings of PAAM, volume 99, page 33. London, 1999

    1999

  26. [33]

    O’Reilly Media, Inc

    Betsy Beyer, Chris Jones, Jennifer Petoff, and Niall Richard Murphy. Site reliability engineering: how Google runs production systems. " O’Reilly Media, Inc.", 2016

    2016

  27. [34]

    Large ai models and their applications: Classification, limitations, and potential solutions

    Jing Bi, Ziqi Wang, Haitao Yuan, Xiankun Shi, Ziyue Wang, Jia Zhang, MengChu Zhou, and Rajkumar Buyya. Large ai models and their applications: Classification, limitations, and potential solutions. Software: Practice and Experience, 55(6):1003–1017, January 2025

    2025

  28. [35]

    https: //wandb.ai/site/

    Weights & Biases.Weights & Biases: The AI developer platform to build AI agents, applications, and models with confidence, 2025. https: //wandb.ai/site/

    2025

  29. [36]

    Service-oriented computing.Computer, 39(3):99–101, 2006

    M Bichier and K-J Lin. Service-oriented computing.Computer, 39(3):99–101, 2006

    2006

  30. [37]

    John Wiley & Sons, 2007

    Rafael H Bordini, Jomi Fred Hübner, and Michael Wooldridge.Pro- gramming multi-agent systems in AgentSpeak using Jason. John Wiley & Sons, 2007

    2007

  31. [38]

    Plans and resource-bounded practical reasoning.Computational intelligence, 4(3):349–355, 1988

    Michael E Bratman, David J Israel, and Martha E Pollack. Plans and resource-bounded practical reasoning.Computational intelligence, 4(3):349–355, 1988

    1988

  32. [39]

    Tropos: An agent-oriented software development methodology.Autonomous Agents and Multi-Agent Systems, 8(3):203– 236, 2004

    Paolo Bresciani, Anna Perini, Paolo Giorgini, Fausto Giunchiglia, and John Mylopoulos. Tropos: An agent-oriented software development methodology.Autonomous Agents and Multi-Agent Systems, 8(3):203– 236, 2004

    2004

  33. [40]

    A complete stit logic for knowledge and action, and some of its applications

    Jan Broersen. A complete stit logic for knowledge and action, and some of its applications. InInternational workshop on declarative agent languages and technologies, pages 47–59. Springer, 2008

    2008

  34. [41]

    A robust layered control system for a mobile robot

    Rodney Brooks. A robust layered control system for a mobile robot. IEEE journal on robotics and automation, 2(1):14–23, 2003

    2003

  35. [42]

    Language models are few-shot learners

    Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. Language models are few-shot learners. Advances in neural information processing systems, 33:1877–1901, 2020

    1901

  36. [43]

    Borg, omega, and kubernetes.Communications of the ACM, 59(5):50–57, 2016

    Brendan Burns, Brian Grant, David Oppenheimer, Eric Brewer, and John Wilkes. Borg, omega, and kubernetes.Communications of the ACM, 59(5):50–57, 2016

    2016

  37. [44]

    Agent oriented analysis using message/uml

    Giovanni Caire, Wim Coulier, Francisco Garijo, Jorge Gomez, Juan Pavon, Francisco Leal, Paulo Chainho, Paul Kearney, Jamie Stark, Richard Evans, et al. Agent oriented analysis using message/uml. InInternational Workshop on Agent-Oriented Software Engineering, pages 119–135. Springer, 2001

    2001

  38. [45]

    An approach for qos-aware service composition based on genetic algorithms

    Gerardo Canfora, Massimiliano Di Penta, Raffaele Esposito, and Maria Luisa Villani. An approach for qos-aware service composition based on genetic algorithms. InProceedings of the 7th Annual Con- ference on Genetic and Evolutionary Computation, GECCO ’05, page 1069–1075, New York, NY , USA, 2005. Association for Computing Machinery

    2005

  39. [46]

    Dynamic role binding in a service oriented architecture

    Humberto Nicolás Castejón and Rolv Bræk. Dynamic role binding in a service oriented architecture. InInternational Conference on 22 Intelligence in Communication Systems, pages 109–122. Springer, 2005

    2005

  40. [47]

    Agentboard: An analytical evaluation board of multi-turn llm agents.Advances in neural information processing systems, 37:74325–74362, 2024

    Ma Chang, Junlei Zhang, Zhihao Zhu, Cheng Yang, Yujiu Yang, Yaohui Jin, Zhenzhong Lan, Lingpeng Kong, and Junxian He. Agentboard: An analytical evaluation board of multi-turn llm agents.Advances in neural information processing systems, 37:74325–74362, 2024

    2024

  41. [48]

    Deep reinforcement learning from human preferences.Advances in neural information processing systems, 30, 2017

    Paul F Christiano, Jan Leike, Tom Brown, Miljan Martic, Shane Legg, and Dario Amodei. Deep reinforcement learning from human preferences.Advances in neural information processing systems, 30, 2017

    2017

  42. [49]

    Mobilevlm v2: Faster and stronger baseline for vision language model

    Xiangxiang Chu, Limeng Qiao, Xinyu Zhang, Shuang Xu, Fei Wei, Yang Yang, Xiaofei Sun, Yiming Hu, Xinyang Lin, Bo Zhang, et al. Mobilevlm v2: Faster and stronger baseline for vision language model. arXiv preprint arXiv:2402.03766, 2024

    Pith/arXiv arXiv 2024

  43. [50]

    Model checking

    Edmund M Clarke. Model checking. InInternational conference on foundations of software technology and theoretical computer science, pages 54–56. Springer, 1997

    1997

  44. [51]

    Edward H. Clarke. Multipart pricing of public goods.Public Choice, 11(1):17–33, 1971

    1971

  45. [52]

    https: //www.jaegertracing.io/

    CNCF.Jaeger: open source, distributed tracing platform, 2025. https: //www.jaegertracing.io/

    2025

  46. [53]

    https://linkerd.io/

    CNCF.Linkerd: Enterprise power without enterprise complexity, 2025. https://linkerd.io/

    2025

  47. [54]

    Business process execution language for web services

    Francisco Curbera, Frank Leymann, Dieter Roller, and Satish R Thatte. Business process execution language for web services. 2003

    2003

  48. [55]

    Embodied question answering

    Abhishek Das, Samyak Datta, Georgia Gkioxari, Stefan Lee, Devi Parikh, and Dhruv Batra. Embodied question answering. InPro- ceedings of the IEEE conference on computer vision and pattern recognition, pages 1–10, 2018

    2018

  49. [56]

    Plug and play language models: A simple approach to controlled text generation

    Sumanth Dathathri, Andrea Madotto, Janice Lan, Jane Hung, Eric Frank, Piero Molino, Jason Yosinski, and Rosanne Liu. Plug and play language models: A simple approach to controlled text generation. arXiv preprint arXiv:1912.02164, 2019

    Pith/arXiv arXiv 1912

  50. [57]

    Cloud-native computing: A survey from the perspective of services.Proceedings of the IEEE, 112(1):12–46, 2025

    Shuiguang Deng, Hailiang Zhao, Binbin Huang, Cheng Zhang, Feiyi Chen, Yinuo Deng, Jianwei Yin, Schahram Dustdar, and Albert Y Zomaya. Cloud-native computing: A survey from the perspective of services.Proceedings of the IEEE, 112(1):12–46, 2025

    2025

  51. [58]

    Acceler- ating containerized service delivery at the network edge, 2025

    Yinuo Deng, Hailiang Zhao, Dongjing Wang, Peng Chen, Wenzhuo Qian, Jianwei Yin, Schahram Dustdar, and Shuiguang Deng. Acceler- ating containerized service delivery at the network edge, 2025

    2025

  52. [59]

    The dmars architecture: A specification of the distributed multi-agent reasoning system.Autonomous Agents and Multi-Agent Systems, 9(1):5–53, 2004

    Mark d’Inverno, Michael Luck, Michael Georgeff, David Kinny, and Michael Wooldridge. The dmars architecture: A specification of the distributed multi-agent reasoning system.Autonomous Agents and Multi-Agent Systems, 9(1):5–53, 2004

    2004

  53. [60]

    The observable mind: Enabling an autonomous agent sharing its conscious contents using a cognitive architecture

    Daqi Dong. The observable mind: Enabling an autonomous agent sharing its conscious contents using a cognitive architecture. In Proceedings of the AAAI Symposium Series, volume 2, pages 172–176, 2023

    2023

  54. [61]

    A survey on in-context learning.arXiv preprint arXiv:2301.00234, 2022

    Qingxiu Dong, Lei Li, Damai Dai, Ce Zheng, Jingyuan Ma, Rui Li, Heming Xia, Jingjing Xu, Zhiyong Wu, Tianyu Liu, et al. A survey on in-context learning.arXiv preprint arXiv:2301.00234, 2022

    Pith/arXiv arXiv 2022

  55. [62]

    Palm-e: An embodied multimodal language model

    Danny Driess, Fei Xia, Mehdi SM Sajjadi, Corey Lynch, Aakanksha Chowdhery, Ayzaan Wahid, Jonathan Tompson, Quan Vuong, Tianhe Yu, Wenlong Huang, et al. Palm-e: An embodied multimodal language model. 2023

    2023

  56. [63]

    Guoguang Du, Kai Wang, Shiguo Lian, and Kaiyong Zhao. Vision- based robotic grasping from object localization, object pose estimation to grasp estimation for parallel grippers: a review.Artificial Intelligence Review, 54(3):1677–1734, 2021

    2021

  57. [64]

    A survey of embodied ai: From simulators to research tasks

    Jiafei Duan, Samson Yu, Hui Li Tan, Hongyuan Zhu, and Cheston Tan. A survey of embodied ai: From simulators to research tasks. IEEE Transactions on Emerging Topics in Computational Intelligence, 6(2):230–244, 2022

    2022

  58. [65]

    Htn planning: Complexity and expressivity

    Kutluhan Erol, James Hendler, and Dana S Nau. Htn planning: Complexity and expressivity. InAAAI, volume 94, pages 1123–1128, 1994

    1994

  59. [66]

    Feature flags and dynamic configuration in microservices

    Dorcas Esther and Elizabeth Oliver. Feature flags and dynamic configuration in microservices. 2025

    2025

  60. [67]

    The many faces of publish/subscribe.ACM computing surveys (CSUR), 35(2):114–131, 2003

    Patrick Th Eugster, Pascal A Felber, Rachid Guerraoui, and Anne- Marie Kermarrec. The many faces of publish/subscribe.ACM computing surveys (CSUR), 35(2):114–131, 2003

    2003

  61. [68]

    Automated repair of programs from large language models

    Zhiyu Fan, Xiang Gao, Martin Mirchev, Abhik Roychoudhury, and Shin Hwei Tan. Automated repair of programs from large language models. In2023 IEEE/ACM 45th International Conference on Software Engineering (ICSE), pages 1469–1481. IEEE, 2023

    2023

  62. [69]

    Agent/group/roles: Simulating with organizations

    Jacques Ferber, Olivier Gutknecht, Fabien Michel, et al. Agent/group/roles: Simulating with organizations. InFourth International Workshop on Agent-Based Simulation (ABS03), 2003

    2003

  63. [70]

    Towards trustworthy ai: A review of ethical and robust large language models.arXiv preprint arXiv:2407.13934, 2024

    Md Meftahul Ferdaus, Mahdi Abdelguerfi, Elias Ioup, Kendall N Niles, Ken Pathak, and Steven Sloan. Towards trustworthy ai: A review of ethical and robust large language models.arXiv preprint arXiv:2407.13934, 2024

    Pith/arXiv arXiv 2024

  64. [71]

    Fipa communicative act library specification.Change, 2000(01/18), 2000

    T FIPA. Fipa communicative act library specification.Change, 2000(01/18), 2000

    2000

  65. [72]

    Magentic-one: A generalist multi-agent system for solving complex tasks.arXiv preprint arXiv:2411.04468, 2024

    Adam Fourney, Gagan Bansal, Hussein Mozannar, Cheng Tan, Eduardo Salinas, Friederike Niedtner, Grace Proebsting, Griffin Bassman, Jack Gerrits, Jacob Alber, et al. Magentic-one: A generalist multi-agent system for solving complex tasks.arXiv preprint arXiv:2411.04468, 2024

    Pith/arXiv arXiv 2024

  66. [73]

    An overview of it service management.Communications of the ACM, 52(5):124–127, 2009

    Stuart D Galup, Ronald Dattero, Jim J Quan, and Sue Conger. An overview of it service management.Communications of the ACM, 52(5):124–127, 2009

    2009

  67. [74]

    Pal: Program-aided language models, 2022

    Luyu Gao, Aman Madaan, Shuyan Zhou, Uri Alon, Pengfei Liu, Yiming Yang, Jamie Callan, and Graham Neubig. Pal: Program-aided language models, 2022

    2022

  68. [75]

    Datasheets for datasets.Communications of the ACM, 64(12):86–92, 2021

    Timnit Gebru, Jamie Morgenstern, Briana Vecchione, Jennifer Wort- man Vaughan, Hanna Wallach, Hal Daumé Iii, and Kate Crawford. Datasheets for datasets.Communications of the ACM, 64(12):86–92, 2021

    2021

  69. [76]

    Modeling rational agents within a bdi- architecture

    M Georgeff and A Rao. Modeling rational agents within a bdi- architecture. InProc. 2nd Int. Conf. on Knowledge Representation and Reasoning (KR’91). Morgan Kaufmann, pages 473–484. of, 1991

    1991

  70. [77]

    Reactive reasoning and planning

    Michael P Georgeff and Amy L Lansky. Reactive reasoning and planning. InAAAI, volume 87, pages 677–682, 1987

    1987

  71. [78]

    An overview of hierarchical task network planning.arXiv preprint arXiv:1403.7426, 2014

    Ilche Georgievski and Marco Aiello. An overview of hierarchical task network planning.arXiv preprint arXiv:1403.7426, 2014

    Pith/arXiv arXiv 2014

  72. [79]

    Decentralized task alloca- tion in multi-robot systems via bipartite graph matching augmented with fuzzy clustering

    Payam Ghassemi and Souma Chowdhury. Decentralized task alloca- tion in multi-robot systems via bipartite graph matching augmented with fuzzy clustering. InInternational design engineering technical conferences and computers and information in engineering conference, volume 51753, page V02AT03A014. American Society of Mechanical Engineers, 2018

    2018

  73. [80]

    Imagebind: One embedding space to bind them all

    Rohit Girdhar, Alaa Ghosh, Kristen Grauman, Thalaiyasingam Ajan- than, Joao Carreira, Christoph Feichtenhofer, and Pedro Pinheiro. Imagebind: One embedding space to bind them all. InIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 20017–20027, 2023

    2023

  74. [81]

    https://cloud.google.com/ products/agent-builder

    Google.Vertex AI Agent Builder, 2025. https://cloud.google.com/ products/agent-builder

    2025

  75. [82]

    Iqa: Visual question answering in interactive environments

    Daniel Gordon, Aniruddha Kembhavi, Mohammad Rastegari, Joseph Redmon, Dieter Fox, and Ali Farhadi. Iqa: Visual question answering in interactive environments. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 4089–4098, 2018

    2018

  76. [83]

    Hybrid computing using a neural network with dynamic external memory

    Alex Graves, Greg Wayne, Malcolm Reynolds, Tim Harley, Ivo Dani- helka, Agnieszka Grabska-Barwi ´nska, Sergio Gómez Colmenarejo, Edward Grefenstette, Tiago Ramalho, John Agapiou, et al. Hybrid computing using a neural network with dynamic external memory. Nature, 538(7626):471–476, 2016

    2016

  77. [84]

    Incentives in teams.Econometrica, 41(4):617–631, 1973

    Theodore Groves. Incentives in teams.Econometrica, 41(4):617–631, 1973

    1973

  78. [85]

    Chawla, Olaf Wiest, and Xiangliang Zhang

    Taicheng Guo, Xiuying Chen, Yaqi Wang, Ruidi Chang, Shichao Pei, Nitesh V . Chawla, Olaf Wiest, and Xiangliang Zhang. Large language model based multi-agents: A survey of progress and challenges, 2025

    2025

  79. [86]

    The off-switch game

    Dylan Hadfield-Menell, Anca D Dragan, Pieter Abbeel, and Stuart Russell. The off-switch game. InAAAI Workshops, 2017

    2017

  80. [87]

    The symbol grounding problem.Physica D: Nonlinear Phenomena, 42(1-3):335–346, 1990

    Stevan Harnad. The symbol grounding problem.Physica D: Nonlinear Phenomena, 42(1-3):335–346, 1990

    1990

Showing first 80 references.