# Self-Compacting Language Model Agents

arXiv: https://arxiv.org/abs/2606.23525 · HTML: https://mlsanity.com/paper/2606.23525/

## Key facts

- Cited by 11 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 33 of its 44 references that are in the map.
- First posted on arXiv 22 Jun 2026.

## Placed in

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

## Builds on (most cited first, 30 of 33)

- [Efficient Memory Management for Large Language Model Serving with PagedAttention](https://mlsanity.com/paper/2309.06180/) (2023)
- [SWE-bench: Can Language Models Resolve Real-World GitHub Issues?](https://mlsanity.com/paper/2310.06770/) (2023)
- [Efficient Streaming Language Models with Attention Sinks](https://mlsanity.com/paper/2309.17453/) (2023)
- [SnapKV: LLM Knows What You are Looking for Before Generation](https://mlsanity.com/paper/2404.14469/) (2024)
- [AlphaEvolve: A coding agent for scientific and algorithmic discovery](https://mlsanity.com/paper/2506.13131/) (2025)
- [BrowseComp: A Simple Yet Challenging Benchmark for Browsing Agents](https://mlsanity.com/paper/2504.12516/) (2025)
- [LLMs Get Lost In Multi-Turn Conversation](https://mlsanity.com/paper/2505.06120/) (2025)
- [MathArena: Evaluating LLMs on Uncontaminated Math Competitions](https://mlsanity.com/paper/2505.23281/) (2025)
- [BrowseComp-Plus: A More Fair and Transparent Evaluation Benchmark of Deep-Research Agent](https://mlsanity.com/paper/2508.06600/) (2025)
- [Scaling Long-Horizon LLM Agent via Context-Folding](https://mlsanity.com/paper/2510.11967/) (2025)
- [SWE-rebench: An Automated Pipeline for Task Collection and Decontaminated Evaluation of Software Engineering Agents](https://mlsanity.com/paper/2505.20411/) (2025)
- [ReSum: Unlocking Long-Horizon Search Intelligence via Context Summarization](https://mlsanity.com/paper/2509.13313/) (2025)
- [A Survey of Context Engineering for Large Language Models](https://mlsanity.com/paper/2507.13334/) (2025)
- [AgentFold: Long-Horizon Web Agents with Proactive Context Management](https://mlsanity.com/paper/2510.24699/) (2025)
- [WebExplorer: Explore and Evolve for Training Long-Horizon Web Agents](https://mlsanity.com/paper/2509.06501/) (2025)
- [Cartridges: Lightweight and general-purpose long context representations via self-study](https://mlsanity.com/paper/2506.06266/) (2025)
- [Emergent Introspective Awareness in Large Language Models](https://mlsanity.com/paper/2601.01828/) (2026)
- [MiroThinker: Pushing the Performance Boundaries of Open-Source Research Agents via Model, Context, and Interactive Scaling](https://mlsanity.com/paper/2511.11793/) (2025)
- [Towards Robust Mathematical Reasoning](https://mlsanity.com/paper/2511.01846/) (2025)
- [InftyThink: Breaking the Length Limits of Long-Context Reasoning in Large Language Models](https://mlsanity.com/paper/2503.06692/) (2025)
- [KeyDiff: Key Similarity-Based KV Cache Eviction for Long-Context LLM Inference in Resource-Constrained Environments](https://mlsanity.com/paper/2504.15364/) (2025)
- [DeepSearchQA: Bridging the Comprehensiveness Gap for Deep Research Agents](https://mlsanity.com/paper/2601.20975/) (2026)
- [Fast KV Compaction via Attention Matching](https://mlsanity.com/paper/2602.16284/) (2026)
- [MEMENTO: Teaching LLMs to Manage Their Own Context](https://mlsanity.com/paper/2604.09852/) (2026)
- [Agentic Aggregation for Parallel Scaling of Long-Horizon Agentic Tasks](https://mlsanity.com/paper/2604.11753/) (2026)
- [Reasoning Cache: Continual Improvement Over Long Horizons via Short-Horizon RL](https://mlsanity.com/paper/2602.03773/) (2026)
- [FoldAct: Efficient and Stable Context Folding for Long-Horizon Search Agents](https://mlsanity.com/paper/2512.22733/) (2025)
- [Accordion-Thinking: Self-Regulated Step Summaries for Efficient and Readable LLM Reasoning](https://mlsanity.com/paper/2602.03249/) (2026)
- [Breadcrumbs Reasoning: Memory-Efficient Reasoning with Compression Beacons](https://mlsanity.com/paper/2510.13797/) (2025)
- [MOTIF: Modular Thinking via Reinforcement Fine-tuning in LLMs](https://mlsanity.com/paper/2507.02851/) (2025)

## Built on by (most cited first, 11 of 11)

- [Prefix Sliding for efficient test-time scaling](https://mlsanity.com/paper/2608.26070/) (2026)
- [Memory Control Signals Emerge Before Action in Long Horizon Agents](https://mlsanity.com/paper/2609.27286/) (2026)
- [StateComp: Learning When to Compress History in Long Horizon Agents](https://mlsanity.com/paper/2609.27298/) (2026)
- [KV-streams for Efficient Compaction in Agentic Reinforcement Learning](https://mlsanity.com/paper/2609.35750/) (2026)
- [Context Language Models](https://mlsanity.com/paper/2609.37725/) (2026)
- [CliffCompaction: Cost-Efficient Compaction for Long-Horizon Coding Agents](https://mlsanity.com/paper/2609.26779/) (2026)
- [DRSR: Learning Set-Level Deletion Risk for Efficient Long-Horizon Agents](https://mlsanity.com/paper/2609.27276/) (2026)
- [Stable Geometry with Divergent Task Evidence for Efficient Long-Horizon Agent Compression](https://mlsanity.com/paper/2609.27332/) (2026)
- [When Can Agents Forget Their Reasoning? ICLR for Long-Horizon Agent Context Compression](https://mlsanity.com/paper/2609.29875/) (2026)
- [Continuous Context Management](https://mlsanity.com/paper/2609.35540/) (2026)
- [AutoCompact: Learning When to Compact Context in Long-Horizon Coding Agents](https://mlsanity.com/paper/2610.02163/) (2026)

## Abstract (arXiv)

Long agent traces composed of chains of thought and tool calls accumulate stale content that anchor subsequent generations, and eventually outgrow the context window. Existing scaffolds mitigate it with fixed-interval compaction triggered at a token threshold. Such triggers pay no heed to trajectory structure, risking discard of partial results mid-derivation or mid-search. We propose SelfCompact, a scaffold that allows the model itself to decide when and how to compact. Specifically, it pairs two inference-time elements: (i) a compaction tool the model invokes to summarize the accumulated context, and (ii) a lightweight rubric specifying when to fire (a sub-task has resolved, or the trajectory is converging) and when to suppress (mid-derivation, or when stuck). Both are needed. The tool alone is unevenly used across open-weight models, often invoked at unhelpful moments or not at all; the rubric alone cannot act. Together, they elicit effective adaptive compaction without any fine-tuning or external supervision. We present empirical results on six benchmarks (competitive math and agentic search) and seven models. Our results show that SelfCompact matches or exceeds fixed-interval summarization at a fraction of the token cost, improving over a no-summarization baseline by up to 18.1 points on math and 5-9 points on agentic search at 30-70% lower per-question cost. Our results expose a meta-cognitive gap: although unprompted models cannot reliably tell when their own context is rotting, a lightweight rubric closes this gap, reframing when to compact as a capability that scaffolds can supply without training.