Who May Overrule the Agent? Evidence-Gated Authority in LLM Review Institutions

Computer Science > Artificial Intelligence arXiv:2611.05390 (cs) [Submitted on 12 Sep 2026] Title: Who May Overrule the Agent? Evidence-Gated Authority in LLM Review Institutions Authors: Fuad Ali.

计算机科学 > 人工智能 arXiv:2611.05390 (cs) [提交于 2026 年 9 月 12 日] 标题：谁可以推翻代理？大语言模型审查机构中的证据门控权限 作者：Fuad Ali。

Abstract: Multi-agent language-model systems often add a reviewer to improve an upstream agent’s decision. Yet a reviewer can both repair errors and replace decisions that were already correct. We study this as a problem of institutional jurisdiction: conditional on the same proposal, task information, model, and review output, when is the reviewer authorized to make its recommendation binding?

摘要：多智能体语言模型系统通常会增加一名审查员来改进上游智能体的决策。然而，审查员既可以修复错误，也可能替换掉原本正确的决策。我们将此作为制度管辖权问题进行研究：在相同的提案、任务信息、模型和审查输出条件下，审查员在何时被授权使其建议具有约束力？

We compare broad override authority with an evidence gate that permits replacement only after the reviewer produces a mechanically valid record of a binding-constraint violation and a compliant repair. Across two locally run open-weight models, 96 spatial-policy tasks, and a 96-task portfolio-transfer benchmark, the gate sharply reduces corruption of correct proposals while retaining most repairs of violations.

我们将广泛的推翻权限与证据门控进行了比较，后者仅在审查员生成了约束违规的机械有效记录及合规修复方案后，才允许进行替换。在两个本地运行的开放权重模型、96 个空间策略任务以及一个 96 任务的投资组合转移基准测试中，该门控机制显著减少了对正确提案的破坏，同时保留了对违规行为的大部分修复。

The directions of this tradeoff are partly implied by the rule; its magnitude depends on reviewer transitions and upstream proposal states. A prospective validation then gives the reviewer an explicit choice to retain, replace, or escalate. This eliminates correct-proposal corruption for Qwen, but Mistral still replaces all 48 correct proposals; Qwen also declines every welfare-only repair that Mistral frequently attempts.

这种权衡的方向部分由规则所暗示；其程度取决于审查员的转换和上游提案的状态。随后，前瞻性验证赋予了审查员保留、替换或升级的明确选择权。这消除了 Qwen 对正确提案的破坏，但 Mistral 仍然替换了所有 48 个正确提案；Qwen 还拒绝了 Mistral 频繁尝试的每一项仅涉及福利的修复。

Thus interface design changes reviewer behavior but is not a model-independent safeguard. The result is not that restricted review is universally better. It is that reviewer competence, action interface, and binding jurisdiction jointly determine both the correction domain and the corruption surface.

因此，界面设计会改变审查员的行为，但并非一种独立于模型的保障措施。结果并非受限审查在普遍意义上更好。而是审查员的能力、操作界面和约束管辖权共同决定了修正领域和破坏面。