拆分智能体削弱政策合规性，关键在信息传递损耗
Governance at the Boundary: How Agent Decomposition Degrades Policy Compliance
- 通过多组件架构测试金融代理治理能力，发现信息在组件间传递时衰减
- 32B模型在编排架构下626次实验中56%~85%的合规事实丢失
- 同一机制既导致漏报风险也引发误报，适合关注AI治理的开发者
现有智能体评估关注任务完成度，本文关注任务是否符合政策。我们提出Fiducia-bench，用于评估金融智能体的可治理性——是否在必要时升级、在要求时回避、并留下可审计痕迹。利用该基准研究了此前未被覆盖的问题：将智能体分解为多个组件是否会削弱其治理能力？结果表明确实会，且机制明确：一个组件发现的关键合规信息，在传给需行动的组件时发生衰减。在100种KYC/AML任务变体、两个模型和三种架构的626次实验中，32B开源模型在单循环基线中仅衰减0%，固定流水线中衰减56%，编排-子代理架构中衰减85%（约束距离为2）。更强模型gpt-4.1-mini在同一条件下衰减3-6%，说明治理代价部分与模型能力相关。关键的是，该机制同时引发低估风险和高估风险，取决于丢失的信息是风险信号还是免责证据。基准、任务和验证工具均已开源。
原文摘要 · Abstract (English)
Existing agent benchmarks ask whether the agent finished the task. We ask whether it finished it within policy. We introduce Fiducia-bench, a benchmark for the governability of financial agents---whether they escalate when obligated, abstain when required, and leave an auditable trail---and use it to study a question no prior benchmark addresses: does decomposing an agent into components degrade its governance? It does, and the mechanism is specific. Policy-relevant facts discovered by one component are attenuated at the handoff boundary before reaching the component that must act on them. In a 626-episode experiment across 100 KYC/AML task variants, two models, and three architectures, a 32B open-weights model attenuated 0% of discovered facts under a single-loop baseline, 56% under a fixed pipeline, and 85% under an orchestrator-subagent architecture (all at constraint distance 2). A stronger model (gpt-4.1-mini) attenuated 3-6% under the same conditions, suggesting the governance cost of decomposition is partly a function of model capability. Critically, the same mechanism produces both under-escalation and over-escalation, depending on whether the dropped fact was a risk signal or an exculpating one. The benchmark, all tasks, and the verification harness are open-source
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。