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

• URL: http://arxiv.org/abs/2608.06503v1
• Date: Thu, 06 Aug 2026 18:42:02 GMT
• Title: Toward Reliable Context Compression for Long-Horizon Agents: An Empirical Study of Execution Instability
• Abstract summary: We show that compression can weaken the influence of recent interactions, increasing blocked actions, repeated exploration, and instability across runs.<n>We introduce TRACE, a verifier-guided framework that evaluates individual compaction events through paired closed-loop continuations from the same environment state.<n>Initial results on AppWorld show improvements over existing compression baselines in task performance, multi-run reliability, and context--execution efficiency.
• Score: 7.183748353637384
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: 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.

Related papers

• CoACT: Action-Preserving Observation Compression for Coding Agents [8.484842324052751]
  This paper proposes CoACT, an action-preserving observation compression method for coding agents.<n>CoACT is built on next-action preservation (NAP), which requires a compressed observation to induce the same next action as the raw observation.<n> Experiments on SWE-bench Verified with three agentic models show that CoACT reduces average total token consumption by 33.0%.
  arXiv  Detail & Related papers  (2026-07-03T03:17:52Z)
• Learning to Watch: Active Video Anomaly Understanding via Interleaved Policy Optimization [56.150184676745205]
  Video anomaly understanding (VAU) relies on sparse, context-dependent cues.<n>We propose $Anomtext-$, a closed-loop framework that reconceptualizes video understanding as an active sequential decision-making process.<n>Inspired by human video-reviewing behavior, this framework unifies internal cognitive reasoning and strategic evidence acquisition into an interleaved policy.
  arXiv  Detail & Related papers  (2026-07-01T08:41:49Z)
• Perception Without Engagement: Dissecting the Causal Discovery Deficit in LMMs [68.58207076756237]
  We introduce ProCauEval, a perturbation-based evaluation protocol that shifts from outcome assessment to mechanism diagnosis.<n>We find that models faithfully perceive video content yet systematically underexploit it during causal reasoning.<n>We propose Anti-Distillation Policy Optimization (ADPO), a reinforcement learning framework built on negative teacher alignment.
  arXiv  Detail & Related papers  (2026-05-10T08:48:58Z)
• Efficient Agent Evaluation via Diversity-Guided User Simulation [7.723545220477047]
  DIVERT is a snapshot-based, coverage-guided user simulation framework for systematic exploration of agent-user interactions.<n>It captures the full agent-environment state at critical decision points and resumes execution from these snapshots.<n>By focusing evaluation on semantically diverse and underexplored trajectories, DIVERT improves both efficiency and coverage.
  arXiv  Detail & Related papers  (2026-04-23T09:41:21Z)
• PoC: Performance-oriented Context Compression for Large Language Models via Performance Prediction [40.705016911274]
  We introduce a paradigm shift to Performance-oriented Context Compression (PoC)<n>PoC employs a lightweight performance predictor to automatically find the most aggressive compression ratio.<n>Our work paves the way for a more reliable, efficient, and performance-aware deployment of context compression for Large Language Models.
  arXiv  Detail & Related papers  (2026-03-20T08:18:46Z)
• On the Structural Non-Preservation of Epistemic Behaviour under Policy Transformation [51.56484100374058]
  We formalise such information-conditioned interaction patterns as behavioural dependency.<n>This induces a probe-relative notion of $$-behavioural equivalence and a within-policy behavioural distance.<n>Results identify structural conditions under which probe-conditioned behavioural separation is not preserved under common policy transformations.
  arXiv  Detail & Related papers  (2026-02-24T22:55:21Z)
• ARC: Active and Reflection-driven Context Management for Long-Horizon Information Seeking Agents [9.76162701959422]
  ARC is a framework for systematically formulate context management.<n>It treats context as a dynamic internal reasoning state during execution.<n>It consistently outperforms passive context compression methods.
  arXiv  Detail & Related papers  (2026-01-17T12:17:50Z)
• ConCISE: Confidence-guided Compression in Step-by-step Efficient Reasoning [64.93140713419561]
  Large Reasoning Models (LRMs) perform strongly in complex reasoning tasks via Chain-of-Thought (CoT) prompting, but often suffer from verbose outputs.<n>Existing fine-tuning-based compression methods either operate post-hoc pruning, risking disruption to reasoning coherence, or rely on sampling-based selection.<n>We introduce ConCISE, a framework designed to generate concise reasoning chains, integrating Confidence Injection to boost reasoning confidence, and Early Stopping to terminate reasoning when confidence is sufficient.
  arXiv  Detail & Related papers  (2025-05-08T01:40:40Z)
• CausalBench: A Large-scale Benchmark for Network Inference from Single-cell Perturbation Data [61.088705993848606]
  We introduce CausalBench, a benchmark suite for evaluating causal inference methods on real-world interventional data. CaulBench incorporates biologically-motivated performance metrics, including new distribution-based interventional metrics.
  arXiv  Detail & Related papers  (2022-10-31T13:04:07Z)
• What do Compressed Large Language Models Forget? Robustness Challenges in Model Compression [68.82486784654817]
  We study two popular model compression techniques including knowledge distillation and pruning. We show that compressed models are significantly less robust than their PLM counterparts on adversarial test sets. We develop a regularization strategy for model compression based on sample uncertainty.
  arXiv  Detail & Related papers  (2021-10-16T00:20:04Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.