aiqbench analysis
This study provides the first systems-level characterization of agent memory, offering a new taxonomy and cost profiling for stateful long-horizon workloads. By benchmarking ten representative systems, the authors reveal how specific design choices impact write and read path costs, leading to actionable recommendations for fleet-scale management and latency optimization. These findings are critical for engineers building scalable agent infrastructures who need to balance memory freshness, construction overhead, and retrieval efficiency.
Agent Memory: Characterization and System Implications of Stateful Long-Horizon Workloads
arXiv:2606.06448v1 Announce Type: new Abstract: LLM agents are increasingly deployed on long-horizon tasks requiring sustained reasoning over extended interaction histories. Realizing this at scale requires agents to persistently store, retrieve, and update their own memory across sessions. A rich ecosystem of agent memory systems has emerged spanning flat retrieval, LLM-mediated extraction, consolidating fact stores, and agentic control flows. Yet, their system-level behavior remains uncharacterized. We present the first systems characterization of agent memory. First, we introduce a system-oriented taxonomy classifying agent memory systems along four axes. Second, we build a phase-aware profiling harness attributing cost to construction, retrieval, and generation. Third, we characterize ten representative systems across two benchmark suites, uncovering how design choices shift cost across the write and read paths. Finally, we derive 10 system recommendations covering construction scheduling, capability floors, amortization via query volume, freshness-latency tradeoffs, and fleet-scale management.
Abstract:LLM agents are increasingly deployed on long-horizon tasks requiring sustained reasoning over extended interaction histories. Realizing this at scale requires agents to persistently store, retrieve, and update their own memory across sessions. A rich ecosystem of agent memory systems has emerged spanning flat retrieval, LLM-mediated extraction, consolidating fact stores, and agentic control flows. Yet, their system-level behavior remains uncharacterized. We present the first systems characterization of agent memory. First, we introduce a system-oriented taxonomy classifying agent memory systems along four axes. Second, we build a phase-aware profiling harness attributing cost to construction, retrieval, and generation. Third, we characterize ten representative systems across two benchmark suites, uncovering how design choices shift cost across the write and read paths. Finally, we derive 10 system recommendations covering construction scheduling, capability floors, amortization via query volume, freshness-latency tradeoffs, and fleet-scale management.
Artificial Intelligence (cs.AI)
arXiv:2606.06448 [cs.AI]
(or arXiv:2606.06448v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2606.06448 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.
Full text under the publisher's open licence · 231 words
Related coverage
Insurance of Agentic AI
arXiv · Jun 6, 2026
RBGA: Residual-Budget Guide Arbitration for Offline-to-Online Safe Reinforcement Learning in Dexterous Manipulation
arXiv · Oct 27, 2026
Bayesian Intelligence from the Outside
arXiv · Sep 16, 2026
[AINews] OpenAI’s Patch the Planet program is 1 of 2 parties to discover PHP2Shell vulnerabilities in core PHP affecting 70% of the web, using GPT-5.6 Sol; coordinated response over 2 months
Latent Space · Oct 22, 2026