何时聚合智能体轨迹可诊断？流量主导的解释与校准弃权（不执行）
When Are Aggregate Agent Traces Diagnosable? Traffic-Governed Interpretation and Calibrated Abstention
浏览论文内容
中文总结 AI 辅助
本研究提出一种流量主导的可诊断性规则，通过参考映射门和匹配运行时门建立暴露，再评分变化，并在轨迹不支持时弃权（不执行），以可靠解释聚合智能体行为。
中文摘要 AI 辅助
运行时轨迹看似透明，但闭环策略决定了哪些状态被访问以及哪些故障变得可见。我们研究了一个模拟酒店定价智能体，该智能体在不同需求机制下将时间、库存和市场状态映射为离散价格动作。当策略很少访问受影响的单元时，故障可能不会留下聚合轨迹。我们将进入仅聚合故障解释视为评分或定位之前的可诊断性决策。一个参考映射门要求重复的干净策略支持；一个匹配的运行时门随后要求干净流和当前流中的联合支持。仅在两者都通过后才进行信号分析。我们在物理组件级别上，在不相交的干净流上校准错误接纳，并通过受影响的干净流量而非名义单元覆盖率来建模检测。在一个冻结的一次性留出集中，55/72（76.4%）个机制-组件单元被参考接纳，代表20个物理组件；54/55通过了匹配的运行时接纳，而被拒绝的单元弃权（不执行）。稳定错误接纳为0/20，单侧精确95%上限为0.1391，满足冻结的0.20标准。在嵌套于这20个簇的540个重复单元臂行中，受影响的干净流量相对于单元覆盖率将负对数似然降低了29.3%，每行增益为0.1264 nats（簇自助95%区间[0.0593, 0.1918]）。添加掩码族及其交互作用将日志损失改善了每行0.0015 nats（单侧上限0.0066），低于冻结的0.01实际充分性裕度。一项开发审计发现，精确最小命中集和贪婪选择在12/12场景中选择了相同的支持，因为单例证据已解决了冲突。结果是一个用于解释聚合智能体行为的有界规则：首先建立暴露，然后对变化进行评分，当轨迹无法支持主张时弃权（不执行）。
英文摘要
Runtime traces can appear transparent, but a closed-loop policy determines which states are visited and which failures become visible. We study a simulated hotel-pricing agent mapping time, inventory, and market state to discrete price actions under varying demand regimes. A fault may leave no aggregate trace when the policy rarely visits affected cells. We treat entry into aggregate-only fault interpretation as a diagnosability decision preceding scoring or localization. A reference-map gate requires repeated clean-policy support; a matched runtime gate then requires joint support in clean and current streams. Signal analysis occurs only after both pass. We calibrate false admission on a disjoint clean stream at the physical-component level and model detection by affected clean traffic rather than nominal cell coverage. In a frozen one-shot heldout, 55/72 (76.4%) regime-component units were reference-admitted, representing 20 physical components; 54/55 passed matched runtime admission, while the rejected unit abstained. Stable false admission was 0/20, with a one-sided exact 95% upper bound of 0.1391, meeting the frozen 0.20 criterion. Across 540 repeated unit-arm rows nested in those 20 clusters, affected clean traffic reduced negative log likelihood by 29.3% relative to cell coverage, a gain of 0.1264 nats per row (cluster-bootstrap 95% interval [0.0593, 0.1918]). Adding mask family and its interaction improved log loss by 0.0015 nats per row (one-sided upper bound 0.0066), below the frozen 0.01 practical-sufficiency margin. A development audit found that exact minimum hitting set and greedy selection chose identical supports in 12/12 scenarios because singleton evidence had resolved the conflicts. The result is a bounded rule for interpreting aggregate agent behavior: first establish exposure, then score change, and abstain when the trace cannot support the claim.
发表机构
- Blossom AI
机构由 AI 辅助整理，请以论文原文为准。
