Shared Selective Persistent Memory for Agentic LLM Systems
Quick Answer
This paper shows that The shared selective persistent memory architecture enhances agentic LLM systems by retaining reusable context while discarding irrelevant traces, achieving a 96% task completion rate in enterprise scenarios, significantly outperforming traditional methods.
Quick Take
The shared selective persistent memory architecture enhances agentic LLM systems by retaining reusable context while discarding irrelevant traces, achieving a 96% task completion rate in enterprise scenarios, significantly outperforming traditional methods. This approach also reduces task time by 14x and token costs by 97x through a zero-token refresh mechanism.
Key Points
- Shared selective persistent memory retains task specs, data schemas, tool configs, and output constraints.
- Achieved 96% task completion rate compared to 79% without memory and 71% with full history.
- Zero-token refresh mechanism eliminates LLM re-invocation for updates, reducing task time by 14x.
- Summary-driven generation cuts token costs by 97x versus raw data injection.
- Naive full-history persistence degrades completion rates by introducing stale traces.
Paper Resources
📖 Reader Mode
~2 min readAbstract:Agentic LLM systems that generate code through multi-turn tool use face a fundamental context problem: each session starts from zero, discarding the configuration choices, domain constraints, data schemas, and tool-use patterns that made previous sessions productive. Naively persisting entire conversation histories is token-inefficient and counterproductive: irrelevant context degrades generation quality. We introduce shared selective persistent memory, an architecture that identifies and retains four categories of reusable context (task specifications, data schemas, tool configurations, and output constraints) while discarding session-specific reasoning traces. Crucially, this memory is shared: workspaces encapsulating selective memory can be transferred across users with role-based access control, enabling collaborative reuse without redundant specification. We implement it in a deployed collaborative workspace platform where LLM agents produce, edit, and maintain git-versioned artifacts (dashboards, reports, and data-driven documents) from heterogeneous sources (CSV, SQL, REST APIs, and MCP servers). A complementary zero-token data refresh mechanism decouples generated programs from runtime data, enabling artifact reuse without re-invocation. Across three enterprise scenarios, shared selective persistent memory achieves 96% task completion (vs. 79% without memory and 71% with full history). Zero-token refresh eliminates LLM re-invocation for recurring updates (14x task-time reduction), while summary-driven generation cuts per-invocation token cost by 97x versus raw data injection. A replication on four public datasets confirms generalizability, with zero-token refresh succeeding in 12/12 trials. Notably, naive full-history persistence actively degrades completion by biasing the agent with stale traces, while selective memory outperforms both extremes.
| Comments: | 11 pages, 2 figures, 4 tables |
| Subjects: | Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA); Software Engineering (cs.SE) |
| ACM classes: | I.2.7 |
| Cite as: | arXiv:2607.09493 [cs.AI] |
| (or arXiv:2607.09493v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2607.09493 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Sanjana Pedada [view email]
[v1]
Fri, 10 Jul 2026 15:07:00 UTC (34 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.AI
See more →Adversarial Social Epistemology for Assemblies of Humans and Large Language Models
The paper introduces Adversarial Social Epistemology (ASE) to analyze how agents manipulate trust in public communications, highlighting mechanisms that undermine the reliability of testimony and inference. It critiques existing frameworks like epistemic bubbles and misinformation diffusion, proposing a new language for understanding trust breaches and auditing inferential chains in densely interactive environments involving humans and large language models.