arXiv cs.CL· Yiwei Zha, Qingyu Meng, Herbert Bos, Min Chen·· 11 小时前AI 评分47
ARCBench：智能体上下文压缩后还能遵守规则吗？
Do Agents Still Follow the Rules? Benchmarking Rule Compliance under Agent Context Compaction
AI 导读
研究者提出 ARCBench，一个基于开源 pi agent harness 的可配置基准，通过工具调用智能体循环从执行动作中测量规则遵守情况。结果显示压缩很少直接删除规则，多数违规发生在规则仍保留在压缩历史中时，智能体把后续请求当作许可而忽略规则；最强的针对压缩器的攻击通过在摘要中篡改规则使违规增加一倍以上，而保留规则的压缩运行违规频率仍是完整历史运行的两倍以上。
正文 · 原文
Abstract:LLM agents compact long histories to fit their context windows, but the compacted context must still carry the operating rules that govern later actions. Prior work shows that violations rise after compaction, but measures compliance outside a complete agent harness and does not identify where each violation comes from. We introduce ARCBench(Agent Rule Compliance Benchmark), a configurable benchmark built on the open-source pi agent harness. Its tool-calling agent loop lets us measure rule compliance from actions taken during execution, bringing the evaluation closer to deployed agent workflows. Using ARCBench, we examine where violations after compaction come from, and how different compaction configurations and adversarial settings change their sources. We find that compaction rarely deletes the rule: most violations after compaction occur while the rule is still in the compacted history, with the agent treating a later request as permission, ignoring the rule, and proceeding anyway. Where violations come from depends on the configuration: how the rule is delivered decides whether the agent would break it even without compaction, while the compaction strategy and budget decide whether the rule survives. The strongest compactor-targeted attack more than doubles violations by corrupting the rule in the summary and changing how the agents behave, whereas the strongest defense restores the rule in most summary but still violates. Finally, retaining a rule does not guarantee compliance: compacted runs that retain the rule violate more than twice as often as runs with the full history.
| Subjects: | Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Computation and Language (cs.CL) |
| Cite as: | arXiv:2611.03223 [cs.CR] |
| (or arXiv:2611.03223v1 [cs.CR] for this version) | |
| https://doi.org/10.48550/arXiv.2611.03223 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Yiwei Zha [view email]
[v1]
Mon, 2 Nov 2026 14:45:36 UTC (146 KB)
来源：arXiv cs.CL · arxiv.org