analiza aiqbench
This paper challenges the prevailing narrative around reinforcement learning in service orchestration by demonstrating that many claimed production failures stem from flawed evaluation methodologies rather than inherent algorithmic weaknesses. The authors reveal a systemic issue where publication incentives prioritize benchmark gains over reproducible operational evidence, suggesting that learned controllers may be more robust to real-world perturbations like telemetry noise than previously documented. For practitioners and researchers in distributed systems, this underscores the critical need for standardized perturbation models and production-grade comparators before scaling RL-based orchestration strategies.
Incentives and Evidence in Learned Service Orchestration
arXiv:2606.16555v1 Announce Type: cross Abstract: Reinforcement learning for service orchestration has been the subject of sustained research for over a decade, yet it is not used in production at scale. The usual explanation is that learned controllers degrade under delayed and noisy telemetry, workload shifts, and uncontrolled tenants. We test whether existing evidence supports that explanation. We evaluate three highly influential RL-based orchestration systems spanning resource allocation, DAG scheduling, and autoscaling, using pre-registered predictions about comparative degradation under production-relevant perturbations and paired inference with family-wise error correction. Across the tests, most predicted performance reversals do not occur. Diagnostic analyses show that these outcomes often reflect comparator collapse, artefact limitations, or evaluation choices rather than evidence that learned controllers tolerate the perturbations. One apparent advantage under observation lag is roughly fortyfold compared to a Kubernetes HPA-equivalent controller. Another widely cited result cannot be reconstructed from its released artefact, and the strongest reproducible margin is far smaller than the published results. Conclusions also reverse under changes in perturbation magnitude and evaluation mode. Based on these results and broader patterns in the literature, we identify an institutional problem. Publication and review incentives favour benchmark gains against convenient comparators, even when those gains provide little evidence of deployment performance. We argue that the problem is not solely technical. Rather, it is institutional, so learned orchestration needs production-grade comparators, registered perturbation models, separate operational metrics, and publication criteria that reward reproducible operational evidence. Without these changes, the literature can grow without establishing whether learning improves orchestration.
Abstract:Reinforcement learning for service orchestration has been the subject of sustained research for over a decade, yet it is not used in production at scale. The usual explanation is that learned controllers degrade under delayed and noisy telemetry, workload shifts, and uncontrolled tenants. We test whether existing evidence supports that explanation. We evaluate three highly influential RL-based orchestration systems spanning resource allocation, DAG scheduling, and autoscaling, using pre-registered predictions about comparative degradation under production-relevant perturbations and paired inference with family-wise error correction. Across the tests, most predicted performance reversals do not occur. Diagnostic analyses show that these outcomes often reflect comparator collapse, artefact limitations, or evaluation choices rather than evidence that learned controllers tolerate the perturbations. One apparent advantage under observation lag is roughly fortyfold compared to a Kubernetes HPA-equivalent controller. Another widely cited result cannot be reconstructed from its released artefact, and the strongest reproducible margin is far smaller than the published results. Conclusions also reverse under changes in perturbation magnitude and evaluation mode. Based on these results and broader patterns in the literature, we identify an institutional problem. Publication and review incentives favour benchmark gains against convenient comparators, even when those gains provide little evidence of deployment performance. We argue that the problem is not solely technical. Rather, it is institutional, so learned orchestration needs production-grade comparators, registered perturbation models, separate operational metrics, and publication criteria that reward reproducible operational evidence. Without these changes, the literature can grow without establishing whether learning improves orchestration.
Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG)
arXiv:2606.16555 [cs.DC]
(or arXiv:2606.16555v1 [cs.DC] for this version)
https://doi.org/10.48550/arXiv.2606.16555 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 346 cuvinte
Subiecte conexe
The Internet of Agentic AI: Communication, Coordination, and Collective Intelligence at Scale
arXiv · 12 iun. 2026
Economy of Minds: Emerging Multi-Agent Intelligence with Economic Interactions
arXiv · 15 oct. 2026
DART: A DAG-Based Reputation and Incentive Framework via Blockchain-Enabled Governance for Trustworthy LLM Multi-Agent Collaboration
arXiv · 9 sept. 2026
AI Coding Agents Get a Stack Overflow of Their Own
InfoQ · 16 iun. 2026