四轴决策对齐用于长周期企业AI代理
Four-Axis Decision Alignment for Long-Horizon Enterprise AI Agents
AI总结：
本文提出四轴对齐框架，用于评估长周期企业AI代理的决策行为，涵盖事实精度、推理连贯性、合规重建和校准回避，通过实验揭示了决策对齐的重要性。
摘要：
长周期企业代理在损失性记忆、多步骤推理和绑定监管约束下做出高风险决策（如贷款审批、索赔审核、临床审查、先决授权）。当前评估报告单一任务成功率指标混淆了不同失败模式，隐藏了代理是否符合其部署环境标准。本文提出长周期决策行为可分解为四个正交对齐轴：事实精度（FRP）、推理连贯性（RCS）、合规重建（CRR）和校准回避（CAR）。CRR是新的监管基础轴；CAR是测量轴，区分覆盖与准确性。在受控基准（LongHorizon-Bench）上进行分解，涵盖贷款资格和保险索赔审核，通过六个记忆架构运行，发现结构聚合准确性无法看到：检索在事实精度上崩溃；以模式为锚点的架构支付了支架税；在事实保留提示下，纯总结是FRP、RCS、EDA和CRR的强基线；所有六个架构在每个案例中都做出决定，暴露了决策对齐轴。分解还揭示了自身预先注册的预测，即总结会失败事实回忆，数据在大规模上逆转了这一点，一个轴级的逆转聚合准确性会隐藏。机构对齐（监管重建）和决策对齐（校准回避）在对齐文献中代表性不足，一旦决定离开实验室，它们就成为承重结构。该框架通过两步转移到任何受监管的决策领域：构建事实模式，校准CRR审计员提示。
英文摘要：
Long-horizon enterprise agents make high-stakes decisions (loan underwriting, claims adjudication, clinical review, prior authorization) under lossy memory, multi-step reasoning, and binding regulatory constraints. Current evaluation reports a single task-success scalar that conflates distinct failure modes and hides whether an agent is aligned with the standards its deployment environment requires. We propose that long-horizon decision behavior decomposes into four orthogonal alignment axes, each independently measurable and failable: factual precision (FRP), reasoning coherence (RCS), compliance reconstruction (CRR), and calibrated abstention (CAR). CRR is a novel regulatory-grounded axis; CAR is a measurement axis separating coverage from accuracy. We exercise the decomposition on a controlled benchmark (LongHorizon-Bench) covering loan qualification and insurance claims adjudication with deterministic ground-truth construction. Running six memory architectures, we find structure aggregate accuracy cannot see: retrieval collapses on factual precision; schema-anchored architectures pay a scaffolding tax; plain summarization under a fact-preservation prompt is a strong baseline on FRP, RCS, EDA, and CRR; and all six architectures commit on every case, exposing a decisional-alignment axis the field has not targeted. The decomposition also surfaced a pre-registered prediction of our own, that summarization would fail factual recall, which the data reversed at large magnitude, an axis-level reversal aggregate accuracy would have hidden. Institutional alignment (regulatory reconstruction) and decisional alignment (calibrated abstention) are under-represented in the alignment literature and become load-bearing once decisions leave the laboratory. The framework transfers to any regulated decisioning domain via two steps: build a fact schema, and calibrate the CRR auditor prompt.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
