Always-On LLM Agents: A Survey of Persistent State, Memory, and Governance Challenges
By
[Submitted on 29 Jun 2026]
Summary
This paper surveys "always-on" LLM agents — systems whose future behavior depends on durable state accumulated across earlier interactions. It treats these agents as persistent-state systems encompassing not just memories but also task ledgers, permissions, credentials, commitments, provenance/audit records, shared state, trigger conditions, and externally committed effects. The survey analyzes the literature through six diagnostic axes (authority, scope, mutability, provenance, recoverability, actionability) and a state lifecycle (write, validate, organize, retrieve, act upon, update, forget, audit, roll back). Based on a 435-work coded corpus, it finds the literature concentrates more on accumulating and retrieving state than on governing, recovering, or relinquishing it. The authors introduce the Always-On Evaluation Protocol (AOEP-v0), a pilot evaluation contract that scores state mutation and recovery obligations rather than answer quality alone, connecting always-on agents to databases, distributed systems, formal methods, capability security, and machine unlearning.
Source
Twitter / XAlways-On LLM Agents: A Survey of Persistent State, Memory, and Governance Challengesarxiv.orgKey quotes
· 5 pulledAlways-on agents are systems whose future behavior depends on durable state accumulated across earlier interactions.
The survey reads the literature through six diagnostic axes for each state item: authority, scope, mutability, provenance, recoverability, and actionability.
Across a 435-work coded corpus... the literature concentrates more heavily on accumulating and retrieving state than on governing, recovering, or relinquishing it.
We therefore introduce the Always-On Evaluation Protocol (AOEP-v0), a pilot evaluation contract that makes these governance requirements concrete by scoring state mutation and recovery obligations rather than answer quality alone.
The resulting agenda connects always-on agents to databases, distributed systems, formal methods, capability security, and machine unlearning.
You might also wanna read
Agent Memory Is Distributed State Management, Not Magic
The article argues that "agent memory" in AI systems is fundamentally just distributed state management rebranded. It draws parallels betwee
Study Reveals "Constraint Decay" in LLM Agents for Backend Code Generation Under Structural Requirements
This paper presents a systematic study on how LLM agents handle structural constraints in multi-file backend code generation. The authors in
arxiv.org·1mo agoAutonomous Agentic Systems: A Practical Guide to "Always-On" Agents
LifeSkill: A Reinforcement Learning Framework for Online Lifelong Learning in LLM Agents
This paper introduces LifeSkill, a two-stage reinforcement learning framework for online lifelong learning in Large Language Model (LLM) age
arxiv.org·11d ago
Redesigning Data Systems for LLM Agent Workloads: Addressing Agentic Speculation Challenges
This research paper argues that data systems need to be redesigned to better support LLM agents, which are predicted to become the dominant
arxiv.org·9mo ago
How LLMs and AI agents are breaking the 20-year-old stateless compute architecture
The article argues that the foundational assumption of modern cloud-native architecture—that state lives in the database while compute is st
zknill.io·1mo ago
Comments
Sign in to join the conversation.
No comments yet. Be the first.