@inproceedings{suleymanov-etal-2026-courtguard,
title = "{C}ourt{G}uard-{PMPD}: Zero-Shot Policy Adaptation in {LLM} Safety without Fine-Tuning using Prompt-Based Modular Policy Database",
author = "Suleymanov, Umid and
Bayramov, Rufiz and
Gafarli, Suad and
Mammadov, Taghi and
Musayeva, Seljan and
Kantarcioglu, Murat",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.1317/",
pages = "27886--27912",
ISBN = "979-8-89176-462-0",
abstract = "Safety judges are typically optimized for static harm definitions, yet real-world deployment policies evolve rapidly. We present , a model-agnostic safety judging framework that enables zero-shot policy adaptation without model fine-tuning. Rather than relying on static weights or vector-search RAG over raw text chunks, CourtGuard-PMPD introduces a system abstraction that compiles external governance documents into a Prompt-Based Modular Policy Database (PMPD). At inference time, an LLM category mapper routes inputs against active policy modules to lightweight label judges. When policies change, the system updates its decision boundary instantly by swapping database modules. Evaluated on MLCommons intra-domain policy adaptation, CourtGuard-PMPD achieves the highest mean accuracy (0.84) and F1 (0.82) across four staged policy removals, outperforming policy-following baselines without retraining. On a policy-aligned 850-example attack suite and an eight-dataset public benchmark suite, it remains highly competitive (0.915 and 0.871 accuracy, respectively). These results position policy-grounded judging as a controllable and competitive alternative to static safety classifiers and benchmark-oriented generic judges."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="suleymanov-etal-2026-courtguard">
<titleInfo>
<title>CourtGuard-PMPD: Zero-Shot Policy Adaptation in LLM Safety without Fine-Tuning using Prompt-Based Modular Policy Database</title>
</titleInfo>
<name type="personal">
<namePart type="given">Umid</namePart>
<namePart type="family">Suleymanov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rufiz</namePart>
<namePart type="family">Bayramov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Suad</namePart>
<namePart type="family">Gafarli</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Taghi</namePart>
<namePart type="family">Mammadov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Seljan</namePart>
<namePart type="family">Musayeva</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Murat</namePart>
<namePart type="family">Kantarcioglu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: EMNLP 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-462-0</identifier>
</relatedItem>
<abstract>Safety judges are typically optimized for static harm definitions, yet real-world deployment policies evolve rapidly. We present , a model-agnostic safety judging framework that enables zero-shot policy adaptation without model fine-tuning. Rather than relying on static weights or vector-search RAG over raw text chunks, CourtGuard-PMPD introduces a system abstraction that compiles external governance documents into a Prompt-Based Modular Policy Database (PMPD). At inference time, an LLM category mapper routes inputs against active policy modules to lightweight label judges. When policies change, the system updates its decision boundary instantly by swapping database modules. Evaluated on MLCommons intra-domain policy adaptation, CourtGuard-PMPD achieves the highest mean accuracy (0.84) and F1 (0.82) across four staged policy removals, outperforming policy-following baselines without retraining. On a policy-aligned 850-example attack suite and an eight-dataset public benchmark suite, it remains highly competitive (0.915 and 0.871 accuracy, respectively). These results position policy-grounded judging as a controllable and competitive alternative to static safety classifiers and benchmark-oriented generic judges.</abstract>
<identifier type="citekey">suleymanov-etal-2026-courtguard</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.1317/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>27886</start>
<end>27912</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T CourtGuard-PMPD: Zero-Shot Policy Adaptation in LLM Safety without Fine-Tuning using Prompt-Based Modular Policy Database
%A Suleymanov, Umid
%A Bayramov, Rufiz
%A Gafarli, Suad
%A Mammadov, Taghi
%A Musayeva, Seljan
%A Kantarcioglu, Murat
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F suleymanov-etal-2026-courtguard
%X Safety judges are typically optimized for static harm definitions, yet real-world deployment policies evolve rapidly. We present , a model-agnostic safety judging framework that enables zero-shot policy adaptation without model fine-tuning. Rather than relying on static weights or vector-search RAG over raw text chunks, CourtGuard-PMPD introduces a system abstraction that compiles external governance documents into a Prompt-Based Modular Policy Database (PMPD). At inference time, an LLM category mapper routes inputs against active policy modules to lightweight label judges. When policies change, the system updates its decision boundary instantly by swapping database modules. Evaluated on MLCommons intra-domain policy adaptation, CourtGuard-PMPD achieves the highest mean accuracy (0.84) and F1 (0.82) across four staged policy removals, outperforming policy-following baselines without retraining. On a policy-aligned 850-example attack suite and an eight-dataset public benchmark suite, it remains highly competitive (0.915 and 0.871 accuracy, respectively). These results position policy-grounded judging as a controllable and competitive alternative to static safety classifiers and benchmark-oriented generic judges.
%U https://aclanthology.org/2026.findings-emnlp.1317/
%P 27886-27912
Markdown (Informal)
[CourtGuard-PMPD: Zero-Shot Policy Adaptation in LLM Safety without Fine-Tuning using Prompt-Based Modular Policy Database](https://aclanthology.org/2026.findings-emnlp.1317/) (Suleymanov et al., Findings 2026)
ACL