---
格式版本: 2
标题: "What Does Context Compression Cost an Agent? Interaction Costs Unrevealed by Task-Completion Metrics"
原文链接: "https://arxiv.org/abs/2608.16370"
发布日期: "2026-08-17"
发布时间校准状态: "found"
发布时间需复核: "否"
发布时间来源: "rule:local:strict_original_body"
发布时间证据: "**\\[v1\\]** Mon, 17 Aug 2026 10:21:36 UTC (138 KB)"
发布时间校准原因: "规则确认唯一严格发布时间，来源 local:strict_original_body"
发布时间校准置信度: "high"
发布时间候选数量: 18
发布时间严格候选数量: 6
发布时间原页读取状态: "source template page reused from URL open"
发布时间未找到原因: ""
发布时间校准时间: "2026-08-20T14:47:35+08:00"
发布时间仲裁状态: "skipped"
发布时间仲裁尝试次数: 0
发布时间仲裁耗时毫秒: 0
发现时间: "2026-08-20T14:38:42+08:00"
入库时间: "2026-08-20T06:47:35.678Z"
来源平台: "arXiv 学术论文搜索"
搜索渠道: "source_template"
搜索词: "https://arxiv.org/search/?query=Oracle&searchtype=all"
匹配关键词:
  - "AI"
相关厂家:
  - "Oracle"
相关专家:
  []
内容类型: "网页"
抓取工具: "Free Fetch + Defuddle"
清洗工具: "Defuddle Markdown + Defuddle/Readability 正文提取"
原始附件:
  []
AI优质: "否"
AI打分: 15
AI分档: "非优质"
AI质检状态: "不通过"
AI打分理由: "论文讨论上下文压缩对智能体交互成本的影响，属于AI算法研究，与超节点/AI Rack/机柜级AI基础设施、供电散热互连等主题完全无关。"
AI质检模型: "ali-deepseek-v4-flash"
AI质检时间: "2026-08-20T14:51:33+08:00"
AI主题相关性: 0
AI来源权威性: 8
AI新颖性: 2
AI技术细节: 0
AI商业部署信号: 0
AI完整性: 5
AI摘要: "该研究提出一种受控运行时测量协议，评估固定24回合工具型agent的上下文压缩重获取成本。结果显示压缩使检索调用显著增加（如GPT-5.5从21.0增至63.9次），但任务完成率无显著变化，表明仅用完成率会掩盖交互成本。"
AI摘要模型: "ali-deepseek-v4-flash"
AI摘要时间: "2026-09-07T03:24:36.965Z"
采集批次: "2026年8月20日14点19分32秒"
采集批次ID: "20260820-141932-079"
去重键: "https://arxiv.org/abs/2608.16370"
---

## Computer Science > Artificial Intelligence

## Title:What Does Context Compression Cost an Agent? Interaction Costs Unrevealed by Task-Completion Metrics

Authors:[Shuyu Liu](https://arxiv.org/search/cs?searchtype=author&query=Liu,+S)

[View PDF](https://arxiv.org/pdf/2608.16370) [HTML (experimental)](https://arxiv.org/html/2608.16370v1)

> Abstract:Task completion is the standard metric for evaluating context compression, yet it is incomplete: compression can increase an agent's interaction cost by forcing it to reacquire dropped state while leaving completion statistically unchanged.  
> We introduce a controlled runtime measurement protocol for reacquisition cost in a bounded-horizon tool-using agent. The agent acts in a deterministic planning environment under a fixed 24-turn horizon. We vary compression severity, compare a dropping operator with a fact-preserving operator, restore dropped state through controlled oracle interventions, and decompose tool calls into retrieval and execution. We evaluate three models across two task regimes.  
> Retrieval calls increase in all six model-regime comparisons and account for almost all added interaction; five of six remain significant after Holm correction. At the prespecified 5x comparison point, completion changes are not significant in any cell. DeepSeek shows a significant completion drop only at 10x compression. GPT-5.5 is the clearest case: completion changes from 80% to 85% (p = 1.0) while retrieval increases from 21.0 to 63.9 calls (p =.002).  
> Retention interventions further separate state quantity, state type, and content validity. Random selection is comparable to an offline hindsight oracle, while replacing retained D-state with semantically irrelevant content increases retrieval by 57% (p <.001) without a significant completion change. In a second environment, ALFWorld, sliding compression produces no retrieval surge, showing that the reacquisition signature is environment-dependent rather than intrinsic to shortening context.  
> Overall, compression can impose hidden interaction costs when execution-relevant state becomes absent and must be reacquired, while completion alone may not expose those costs.

| Subjects: | Artificial Intelligence (cs.AI) |
| --- | --- |
| Cite as: | [arXiv:2608.16370](https://arxiv.org/abs/2608.16370) \[cs.AI\] |
|  | (or [arXiv:2608.16370v1](https://arxiv.org/abs/2608.16370v1) \[cs.AI\] for this version) |
|  | [https://doi.org/10.48550/arXiv.2608.16370](https://doi.org/10.48550/arXiv.2608.16370) |

## Submission history

From: Shuyu Liu \[[view email](https://arxiv.org/show-email/c277cdca/2608.16370)\]  
**\[v1\]** Mon, 17 Aug 2026 10:21:36 UTC (138 KB)

[Which authors of this paper are endorsers?](https://arxiv.org/auth/show-endorsers/2608.16370) | Disable MathJax ([What is MathJax?](https://info.arxiv.org/help/mathjax.html))