mlsanityData to 14 Oct 2026

Toward Reliable Context Compression for Long-Horizon Agents: An Empirical Study of Execution Instability

arXiv 2608.06503 ↗First posted 6 Aug 2026cs.LGcited by 10

On the map: Efficient LLM inference and serving → Efficient reasoning and agent inference

Abstract (arXiv)

Recurrent context compression controls context growth in long-horizon agents, but its behavioral effects remain poorly understood. In this preliminary empirical study, we show that compression can weaken the influence of recent interactions, increasing blocked actions, repeated exploration, and instability across runs. Motivated by these observations, we introduce TRACE, a verifier-guided framework that evaluates individual compaction events through paired closed-loop continuations from the same environment state and uses summary preferences to optimize a natural-language compression prompt while keeping all models frozen. Initial results on AppWorld show improvements over existing compression baselines in task performance, multi-run reliability, and context--execution efficiency. These findings provide early evidence for boundary-local evaluation as a promising direction for reliable agent context compression.

Key facts. Cited by 10 papers in the map (data through 14 Oct 2026). Placed in Efficient reasoning and agent inference, one of 12 directions in Efficient LLM inference and serving. Builds on 22 of its 66 references that are in the map. First posted on arXiv 6 Aug 2026.

Builds on

22 of 66 references.

Closest ancestors · 2

Same direction, newest first

  1. 2026Fix the Structural Bottleneck: Context Compression via Explicit Information Transmission3 Feb
  2. 2025ACON: Optimizing Context Compression for Long-horizon LLM Agents1 Oct

Other references · 20

  1. 2022ReAct: Synergizing Reasoning and Acting in Language Modelscited by 8,695
  2. 2023Reflexion: Language Agents with Verbal Reinforcement Learningcited by 4,276
  3. 2023WebArena: A Realistic Web Environment for Building Autonomous Agentscited by 2,006
  4. 2024τ-bench: A Benchmark for Tool-Agent-User Interaction in Real-World Domainscited by 1,402
  5. 2023MemGPT: Towards LLMs as Operating Systemscited by 1,389
  6. 2025τ2-Bench: Evaluating Conversational Agents in a Dual-Control Environmentcited by 627
  7. 2024Executable Code Actions Elicit Better LLM Agentscited by 564
  8. 2024AppWorld: A Controllable World of Apps and People for Benchmarking Interactive Coding Agentscited by 401
  9. 2023Learning to Compress Prompts with Gist Tokenscited by 390
  10. 2023Adapting Language Models to Compress Contextscited by 356
  11. 2024LLMLingua-2: Data Distillation for Efficient and Faithful Task-Agnostic Prompt Compressioncited by 348
  12. 2023LLMLingua: Compressing Prompts for Accelerated Inference of Large Language Modelscited by 287
8 more
  1. 2023In-context Autoencoder for Context Compression in a Large Language Modelcited by 199
  2. 2023Compressing Context to Enhance Inference Efficiency of Large Language Modelscited by 189
  3. 2025ReSum: Unlocking Long-Horizon Search Intelligence via Context Summarizationcited by 141
  4. 2024Long Context Compression with Activation Beaconcited by 104
  5. 2024OfficeBench: Benchmarking Language Agents across Multiple Applications for Office Automationcited by 70
  6. 2025Scaling LLM Multi-turn RL with End-to-end Summarization-based Context Managementcited by 51
  7. 2024500xCompressor: Generalized Prompt Compression for Large Language Modelscited by 47
  8. 2024TACO-RL: Task Aware Prompt Compression Optimization with Reinforcement Learningcited by 22

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 10.

10 papers.

  1. 2026AnyJev Technical Reportcited by 1
  2. 2026Correct Now, Insufficient Later: Auditing Update Sufficiency in Context Compressioncited by 0
  3. 2026Evaluating Budgeted Context Projection with Unexecuted Companion Runscited by 0
  4. 2026Beyond Token Savings: A Systematic Study of Context Compression in LLM Agentscited by 0
  5. 2026Beyond Skill Evolution: Self-Evolving Context Management Policies for Long-Horizon Agent Harnessescited by 0
  6. 2026Does an Agent's History Tell You When Compaction Will Hurt? A Modest, Bounded Effect on the TRACE Paired-Replay Corpuscited by 0
  7. 2026State Representation Shapes Long-Horizon Agent Learning and Transfer: An Empirical Studycited by 0
  8. 2026Glasseek: A Transparent Recipe for Training and Trustworthy Evaluation of DeepSearch Agentscited by 0
  9. 2026DeLoop: Repeated Actions as Self-Supervision for Long-Horizon Agentscited by 0
  10. 2026Maximising Transparency in Context Compaction Improves Retention of Facts and Commitments for Lightweight Coding Agentscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Adapting Context Compression for Long-Horizon Agents with Counterfactual Continuations
  2. 2026StateComp: Learning When to Compress History in Long Horizon Agents
  3. 2026FOCUS: Training-Free Decision-Preserving Context Compression for LLM Agents
  4. 2026Control Under Compression: Reliability Frontiers for Tool-Using Agents
  5. 2026ACM: Agentic Context Management for Long Horizon Tasks
  6. 2026Experience Compression Spectrum: Unifying Memory, Skills, and Rules in LLM Agents
  7. 2026Beyond Compaction: Structured Context Eviction for Long-Horizon Agents
  8. 2026What Does Context Compression Cost an Agent? Interaction Costs Unrevealed by Task-Completion Metrics