arXiv cs.CL· Anthony Rhodes, Gadi Singer·· 11 小时前AI 评分38
BMMS：面向策略条件的语言模型行为合规度量套件
BMMS: Policy-Conditioned Measurement of Behavioral Compliance in Language Models
AI 导读
研究者提出 Behavioral Misalignment Metric Suite（BMMS），以策略语料为输入、输出合规画像，核心指标 CAS 衡量严重度加权合规达成率，CRM 衡量对抗压力下合规存留比例。
正文 · 原文
Abstract:Behavioral evaluation of language models fixes its dimensions in advance: hallucination, sycophancy, toxicity, over-refusal. An operator in a regulated industry needs a different question answered: does a system do what a particular rulebook requires? We introduce the Behavioral Misalignment Metric Suite (BMMS), which measures policy-conditioned misalignment by taking a policy corpus as an argument and returning a compliance profile over it, for any published policy and any system that answers in natural language. The Compliance Attainment Score (CAS) measures severity-weighted attainment over the obligations that bind the system, and the Compliance Robustness Margin (CRM) the share of that attainment surviving graded adversarial pressure; Behavioral Invariance (BIV) and the Behavioral Disposition Index (BDI) are specified and released with the suite. Dimensions enter only by passing a published admission test whose binding criterion is attribution: a quantity that moves when the judge changes, holding the target fixed, is measuring the instrument. Applying it excluded one of the quantities the suite itself reported. Across 1,003 obligations from eleven documents in five registers, the admission criterion holds 94.4% precision against blinded human judgment on vendor specifications and 50.0% on analytical guidance, extraction fidelity is 94.6%, and re-scoring under judges from two other providers leaves target ordering invariant while the absolute score of a single target spans up to 0.22. We apply it to the two most detailed published behavioral specifications. Own-specification advantage is not resolved from zero once document difficulty is subtracted, but absolute self-compliance is another matter: GPT-4o fails 70.7% of the hard constraints binding it in the Model Spec published by OpenAI, where the two Claude models fail 0.0% and 9.1% of theirs.
| Comments: | 21 pages, 4 figures. NeurIPS 2026 Workshop on Trust-AI-Eval (TAE). Artifact: this https URL |
| Subjects: | Artificial Intelligence (cs.AI); Computation and Language (cs.CL) |
| Cite as: | arXiv:2611.09147 [cs.AI] |
| (or arXiv:2611.09147v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2611.09147 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Anthony Rhodes [view email]
[v1]
Fri, 6 Nov 2026 08:24:58 UTC (171 KB)
来源：arXiv cs.CL · arxiv.org