---
title: "Follow the Policy, Fail the Benchmark: Auditing and Revising Policy-Based Decision Benchmarks"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2611.07042v1"
permalink: "https://feed.astralor.com/items/itm_5f1788d970acb72da8387b65"
published: "2026-11-04T19:04:55.000Z"
score: 57
channel: "ai"
tags: ["多模态","Follow","Policy","Fail","Benchmark","Auditing","Revising Policy-Based Decision","Benchmarks Policy-based"]
---

# Follow the Policy, Fail the Benchmark: Auditing and Revising Policy-Based Decision Benchmarks

## 速览

Policy-based decision benchmarks evaluate whether systems apply written rules to financial and operational cases. Yet a rule-following response can be marked wrong when the benchmark misstates the required decision. We …

## 关键点

- Policy-based decision benchmarks evaluate whether systems apply written rules to financial and operational cases.
- Yet a rule-following response can be marked wrong when the benchmark misstates the required decision.
- We audit nine public policy-based decision benchmarks through source tracing, program execution, and controlled checks.

## 全文

Policy-based decision benchmarks evaluate whether systems apply written rules to financial and operational cases. Yet a rule-following response can be marked wrong when the benchmark misstates the required decision. We audit nine public policy-based decision benchmarks through source tracing, program execution, and controlled checks. We find unsupported targets, reference errors that survive replay or propagate as solved steps, and evaluators that reject acceptable answers or credit incomplete decisions. To measure the scoring impact, we rescore unchanged model responses. On the 96 LegalBench SARA numeric cases, correcting the scorer raises acceptance by more than 80 percentage points for each of three models, without changing references or responses. In RuleArena Tax, restoring a required zero floor changes 27 of 300 references and narrows an observed model gap. A controlled federal-exemption task with explicit ownership and eligibility assumptions also shows that errors in supplied intermediate answers can affect final decisions. We present BAR (Benchmark Audit and Revision), a versioned collection of audit evidence and revisions for reevaluation.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_5f1788d970acb72da8387b65