mlsanityData to 14 Oct 2026

Active Context Compression: Autonomous Memory Management in LLM Agents

arXiv 2601.07190 ↗First posted 12 Jan 2026cs.AIcited by 16

On the map: LLM reasoning → Reasoning for agents, planning and code

also in 2 more
Abstract (arXiv)

Large Language Model (LLM) agents struggle with long-horizon software engineering tasks due to "Context Bloat." As interaction history grows, computational costs explode, latency increases, and reasoning capabilities degrade due to distraction by irrelevant past errors. Existing solutions often rely on passive, external summarization mechanisms that the agent cannot control. This paper proposes Focus, an agent-centric architecture inspired by the biological exploration strategies of Physarum polycephalum (slime mold). The Focus Agent autonomously decides when to consolidate key learnings into a persistent "Knowledge" block and actively withdraws (prunes) the raw interaction history. Using an optimized scaffold matching industry best practices (persistent bash + string-replacement editor), we evaluated Focus on N=5 context-intensive instances from SWE-bench Lite using Claude Haiku 4.5. With aggressive prompting that encourages frequent compression, Focus achieves 22.7% token reduction (14.9M -> 11.5M tokens) while maintaining identical accuracy (3/5 = 60% for both agents). Focus performed 6.0 autonomous compressions per task on average, with token savings up to 57% on individual instances. We demonstrate that capable models can autonomously self-regulate their context when given appropriate tools and prompting, opening pathways for cost-aware agentic systems without sacrificing task performance.

Key facts. Cited by 16 papers in the map (data through 14 Oct 2026). Placed in Reasoning for agents, planning and code, one of 17 directions in LLM reasoning. Builds on 6 of its 9 references that are in the map. First posted on arXiv 12 Jan 2026.

Builds on

6 of 9 references.

Closest ancestors · 1

Same direction, newest first

  1. 2023Language Agent Tree Search Unifies Reasoning Acting and Planning in Language Models6 Oct

Other references · 5

  1. 2023Reflexion: Language Agents with Verbal Reinforcement Learningcited by 4,276
  2. 2023Lost in the Middle: How Language Models Use Long Contextscited by 3,455
  3. 2023Voyager: An Open-Ended Embodied Agent with Large Language Modelscited by 2,191
  4. 2023Efficient Streaming Language Models with Attention Sinkscited by 2,186
  5. 2023MemGPT: Towards LLMs as Operating Systemscited by 1,389

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 16.

16 papers.

Sort:

By area:

  1. 2026The Missing Memory Hierarchy: Demand Paging for LLM Context Windowscited by 8
  2. 2026State Contamination in Memory-Augmented LLM Agentscited by 5
  3. 2026Mitigating Provenance-Role Collapse in Long-Term Agents via Typed Memory Representationcited by 5
  4. 2026CoACT: Action-Preserving Observation Compression for Coding Agentscited by 4
  5. 2026LLM Agents Are Latent Context Managers: Eliciting Self-Managed Context via State Proprioceptioncited by 3
  6. 2026View-oriented Conversation Compiler for Agent Trace Analysiscited by 2
  7. 2026AI Agents Do Not Fail Alone:The Context Fails Firstcited by 2
  8. 2026The Workload-Router-Pool Architecture for LLM Inference Optimization: A Vision Paper from the vLLM Semantic Router Projectcited by 1
  9. 2026ObjectGraph: From Document Injection to Knowledge Traversal -- A Native File Format for the Agentic Eracited by 1
  10. 2026Terminus-4B: Can a Smaller Model Replace Frontier LLMs at Agentic Execution Tasks?cited by 1
  11. 2026Memory Control Signals Emerge Before Action in Long Horizon Agentscited by 1
  12. 2026Correctness-Aware Repository Filtering Under Maximum Effective Context Window Constraintscited by 0
4 more
  1. 2026Compact-Memory LLM Agents via Online Max-Member Clustering and Atom-Aware Packingcited by 0
  2. 2026CliffCompaction: Cost-Efficient Compaction for Long-Horizon Coding Agentscited by 0
  3. 2026When Can Agents Forget Their Reasoning? ICLR for Long-Horizon Agent Context Compressioncited by 0
  4. 2026Glasseek: A Transparent Recipe for Training and Trustworthy Evaluation of DeepSearch Agentscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026FOCUS: Training-Free Decision-Preserving Context Compression for LLM Agents
  2. 2026ACM: Agentic Context Management for Long Horizon Tasks
  3. 2025ACON: Optimizing Context Compression for Long-horizon LLM Agents
  4. 2026Escaping the Context Bottleneck: Active Context Curation for LLM Agents via Reinforcement Learning
  5. 2026GenericAgent: A Token-Efficient Self-Evolving LLM Agent via Contextual Information Density Maximization (V1.0)
  6. 2026Toward Reliable Context Compression for Long-Horizon Agents: An Empirical Study of Execution Instability
  7. 2026PRO-LONG: Programmatic Memory Enables Long-Horizon Reasoning
  8. 2026SimpleMem: Efficient Lifelong Memory for LLM Agents