Shared Selective Memory Boosts Agentic LLM System Performance.

Sanjana Pedada, Aditya Dhavala, Neelraj Patil· July 13, 2026 View original

Summary

This paper introduces shared selective persistent memory for agentic LLM systems, which retains reusable context like task specifications and data schemas while discarding irrelevant session history. This architecture significantly improves task completion, reduces token costs, and enables collaborative reuse in code-generating agents.

Agentic LLM systems that generate code through multi-turn interactions often struggle with a fundamental context problem: each new session effectively starts from scratch, losing valuable configuration choices, domain constraints, and tool-use patterns from previous interactions. Simply retaining full conversation histories is inefficient and can degrade performance by introducing irrelevant information. A new architecture, "shared selective persistent memory," addresses this by intelligently identifying and retaining four key categories of reusable context: task specifications, data schemas, tool configurations, and output constraints. Crucially, this memory is shareable across users with role-based access control, fostering collaborative reuse without redundant setup. Implemented in a collaborative workspace platform where LLM agents create and maintain git-versioned artifacts, this system achieved a 96% task completion rate, significantly outperforming systems without memory (79%) or with full history (71%). A complementary "zero-token data refresh" mechanism further reduces task time by 14x for recurring updates, and summary-driven generation cuts per-invocation token costs by 97x compared to raw data injection. Replication on public datasets confirmed generalizability, highlighting that naive full-history persistence can actively harm agent performance.

Why it matters

Professionals building or using agentic LLM systems can achieve dramatically improved efficiency, higher task completion rates, and reduced operational costs by implementing intelligent memory management and context sharing.

How to implement this in your domain

  1. 1Analyze current agentic LLM workflows for context management inefficiencies and redundant information.
  2. 2Design and implement selective memory mechanisms to retain only critical, reusable context.
  3. 3Explore creating shared workspaces for LLM agents to enable collaborative reuse of configurations and schemas.
  4. 4Develop "zero-token data refresh" strategies to decouple generated programs from runtime data for recurring tasks.
  5. 5Evaluate the impact of selective memory on task completion rates and token costs in your LLM applications.

Who benefits

Software DevelopmentData ScienceBusiness IntelligenceIT OperationsConsulting

Key takeaways

  • Selective persistent memory significantly boosts agentic LLM performance and task completion.
  • Sharing reusable context across users enhances collaboration and reduces redundancy.
  • Zero-token data refresh drastically cuts re-invocation time for recurring updates.
  • Naive full-history persistence can degrade agent performance.

Original post by Sanjana Pedada, Aditya Dhavala, Neelraj Patil

"arXiv:2607.09493v1 Announce Type: new Abstract: 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…"

View on X

Originally posted by Sanjana Pedada, Aditya Dhavala, Neelraj Patil on X · view source

Want to go deeper?

Turn these trends into skills with Learnijoy's hands-on AI & tech courses.

Explore courses