压缩即认知失败：Agentic LLM 工具如何将已终止进程伪造成已确认结果
Compaction as Epistemic Failure: How Agentic LLM Tools Fabricate Confirmed Results from Killed Processes
AI 导读
一篇 arXiv 论文记录了 Claude Code 的一种失败模式：超时命令（exit code 143）的部分标准输出被写进压缩摘要，作为已确认结果被后续会话继承，跨会话和模型版本传播假阳性且不再复核。
正文 · 原文
Abstract:Agentic LLM coding tools compress long session histories into compaction summaries that subsequent sessions inherit as ground truth. This paper documents a failure mode in Claude Code where partial standard output from timed-out commands (exit code 143) is recorded in compaction summaries as confirmed results, propagating false positives across sessions and model versions without re-verification. The underlying mechanism is a conflation of observation and persistence, where information that appeared in the terminal is treated as equivalent to information written to durable storage. This finding extends the analysis of LLM self-evaluation failures reported in prior work on non-determinism in LLM-as-judge grading by showing that agentic tools exhibit analogous reliability deficits when reporting on their own operational outcomes. The failure has direct implications for any workflow that relies on agentic session continuity for data processing, scientific computation, or multi-step automation.
| Comments: | 8 pages, companion to arXiv:2606.26185 |
| Subjects: | Software Engineering (cs.SE); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2607.13071 [cs.SE] |
| (or arXiv:2607.13071v2 [cs.SE] for this version) | |
| https://doi.org/10.48550/arXiv.2607.13071 arXiv-issued DOI via DataCite |
Submission history
From: Hiroki Tamba [view email]
[v1]
Sat, 11 Jul 2026 07:45:13 UTC (11 KB)
[v2]
Sat, 24 Oct 2026 07:40:06 UTC (11 KB)
来源：arXiv cs.AI · arxiv.org