# The Compaction Cliff in Long-Running AI Agent Memory

arXiv: https://arxiv.org/abs/2608.22752 · HTML: https://mlsanity.com/paper/2608.22752/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Agent memory systems, one of 16 directions in LLM agents and tool use.
- Builds on 28 of its 41 references that are in the map.
- First posted on arXiv 24 Aug 2026.

## Placed in

- LLM agents and tool use / [Agent memory systems](https://mlsanity.com/agents/agent-memory-systems/)

## Builds on (most cited first, 28 of 28)

- [Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks](https://mlsanity.com/paper/2005.11401/) (2020)
- [Lost in the Middle: How Language Models Use Long Contexts](https://mlsanity.com/paper/2307.03172/) (2023)
- [Efficient Streaming Language Models with Attention Sinks](https://mlsanity.com/paper/2309.17453/) (2023)
- [Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection](https://mlsanity.com/paper/2310.11511/) (2023)
- [Jailbroken: How Does LLM Safety Training Fail?](https://mlsanity.com/paper/2307.02483/) (2023)
- [LongBench: A Bilingual, Multitask Benchmark for Long Context Understanding](https://mlsanity.com/paper/2308.14508/) (2023)
- [τ-bench: A Benchmark for Tool-Agent-User Interaction in Real-World Domains](https://mlsanity.com/paper/2406.12045/) (2024)
- [What Disease does this Patient Have? A Large-scale Open Domain Question Answering Dataset from Medical Exams](https://mlsanity.com/paper/2009.13081/) (2020)
- [RULER: What's the Real Context Size of Your Long-Context Language Models?](https://mlsanity.com/paper/2404.06654/) (2024)
- [BEIR: A Heterogenous Benchmark for Zero-shot Evaluation of Information Retrieval Models](https://mlsanity.com/paper/2104.08663/) (2021)
- [A-MEM: Agentic Memory for LLM Agents](https://mlsanity.com/paper/2502.12110/) (2025)
- [LongMemEval: Benchmarking Chat Assistants on Long-Term Interactive Memory](https://mlsanity.com/paper/2410.10813/) (2024)
- [Evaluating the Factual Consistency of Abstractive Text Summarization](https://mlsanity.com/paper/1910.12840/) (2019)
- [The Instruction Hierarchy: Training LLMs to Prioritize Privileged Instructions](https://mlsanity.com/paper/2404.13208/) (2024)
- [LLMLingua-2: Data Distillation for Efficient and Faithful Task-Agnostic Prompt Compression](https://mlsanity.com/paper/2403.12968/) (2024)
- [R-Judge: Benchmarking Safety Risk Awareness for LLM Agents](https://mlsanity.com/paper/2401.10019/) (2024)
- [RECOMP: Improving Retrieval-Augmented LMs with Compression and Selective Augmentation](https://mlsanity.com/paper/2310.04408/) (2023)
- [xRAG: Extreme Context Compression for Retrieval-augmented Generation with One Token](https://mlsanity.com/paper/2405.13792/) (2024)
- [How Memory Management Impacts LLM Agents: An Empirical Study of Experience-Following Behavior](https://mlsanity.com/paper/2505.16067/) (2025)
- [PersonaRAG: Enhancing Retrieval-Augmented Generation Systems with User-Centric Agents](https://mlsanity.com/paper/2407.09394/) (2024)
- [Forgetful but Faithful: A Cognitive Memory Architecture and Benchmark for Privacy-Aware Generative Agents](https://mlsanity.com/paper/2512.12856/) (2025)
- [Graph-Native Cognitive Memory for AI Agents: Formal Belief Revision Semantics for Versioned Memory Architectures](https://mlsanity.com/paper/2603.17244/) (2026)
- [An Information Theoretic Perspective on Agentic System Design](https://mlsanity.com/paper/2512.21720/) (2025)
- [Adaptive Focus Memory for Language Models](https://mlsanity.com/paper/2511.12712/) (2025)
- [QUITO-X: A New Perspective on Context Compression from the Information Bottleneck Theory](https://mlsanity.com/paper/2408.10497/) (2024)
- [Hierarchical Memory Orchestration for Personalized Persistent Agents](https://mlsanity.com/paper/2604.01670/) (2026)
- [OwlerLite: Scope- and Freshness-Aware Web Retrieval for LLM Assistants](https://mlsanity.com/paper/2601.17824/) (2026)
- [Metadata, Structure, or Strategy? A Decomposition of RAG Context Enrichment](https://mlsanity.com/paper/2606.29645/) (2026)

## Built on by (most cited first, 1 of 1)

- [Correct Now, Insufficient Later: Auditing Update Sufficiency in Context Compression](https://mlsanity.com/paper/2609.20045/) (2026)

## Abstract (arXiv)

A safety rule and an episodic log compete for the same tokens in an AI agent's context. When the budget overflows, both are summarized at the same rate; only the rule needs exact wording to remain enforceable. On 20 production agent configurations, Claude Code's /compact prompt on Sonnet 4.6 preserves 53% of safety rules after one compaction round and 10% after five. We name this the Compaction Cliff. We address it with Knowledge Triage, a framework that classifies each line of an agent's knowledge base by type and routes each type through its own retention policy. Three deterministic operators implement this triage across the three context-management operations: TypeCompact rewrites items in place under per-type fidelity, TypeDecompose partitions a topic too large to compact safely, replicating in-scope safety rules across partitions, and TypeRetrieve fetches items from external storage with in-scope rules pinned ahead of relevance. On five public corpora, TypeCompact preserves 2--4× more safety rules than the strongest single-shot LLM compactor at every ratio, with 96% recall over five rounds. TypeDecompose reaches 0% locality violations against 93% under uniform partitioning. TypeRetrieve reaches 100% recall@50 against 73% for the best single-shot LLM retriever. On three downstream behavioral benchmarks, we outperform the production Sonnet compactor on medical compliance (paired McNemar p < 10-8 on preservation, N = 200), the full-policy and hierarchical baselines on retail task pass rate (p < 0.01, N = 115), and the hierarchical compaction on the airline domain (p = 0.024). We release AgentArtifactCorpus (396,934 agent configurations from 54,628 public GitHub repositories), the classifier, and the reference implementation.