@inproceedings{ma-2026-prc,
title = "{PRC}: An Efficient and Unified Framework for Evaluating Robustness and Controllability of Large Language Models on Few-shot Prompt Variations",
author = "Ma, Chunpeng",
editor = "Zhang, Yue and
Xu, Ruifeng and
Purwarianti, Ayu and
Rong, Tong and
Suzuki, Jun",
booktitle = "Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the {A}sia-Pacific Chapter of the {A}ssociation for {C}omputational {L}inguistics (Volume 1: Long Papers)",
month = nov,
year = "2026",
address = "Hengqin, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.aacl-long.10/",
pages = "204--222",
ISBN = "979-8-89176-439-2",
abstract = "Understanding the impact of prompt variations on large language model (LLM) outputs is critical. Despite extensive research, two challenges persist: conflated terminology stemming from loosely defined concepts, and inefficient evaluations reliant on output generation. To address these, we first formalize prompt categorization to clearly define prompt robustness and controllability. We then introduce the prompt vector (PV), a compact internal representation that reformulates these properties at the representation level. Building on PV, we propose PRC, an efficient and unified framework to systematically evaluate LLMs under few-shot prompt variations without requiring costly autoregressive decoding. Experiments across diverse instruction-tuned LLMs demonstrate that PRC tightly correlates with traditional output-based metrics while substantially reducing computational overhead. Overall, PRC offers a principled and highly efficient tool for diagnosing LLM behavior."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="ma-2026-prc">
<titleInfo>
<title>PRC: An Efficient and Unified Framework for Evaluating Robustness and Controllability of Large Language Models on Few-shot Prompt Variations</title>
</titleInfo>
<name type="personal">
<namePart type="given">Chunpeng</namePart>
<namePart type="family">Ma</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yue</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ruifeng</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ayu</namePart>
<namePart type="family">Purwarianti</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tong</namePart>
<namePart type="family">Rong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jun</namePart>
<namePart type="family">Suzuki</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Hengqin, China</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-439-2</identifier>
</relatedItem>
<abstract>Understanding the impact of prompt variations on large language model (LLM) outputs is critical. Despite extensive research, two challenges persist: conflated terminology stemming from loosely defined concepts, and inefficient evaluations reliant on output generation. To address these, we first formalize prompt categorization to clearly define prompt robustness and controllability. We then introduce the prompt vector (PV), a compact internal representation that reformulates these properties at the representation level. Building on PV, we propose PRC, an efficient and unified framework to systematically evaluate LLMs under few-shot prompt variations without requiring costly autoregressive decoding. Experiments across diverse instruction-tuned LLMs demonstrate that PRC tightly correlates with traditional output-based metrics while substantially reducing computational overhead. Overall, PRC offers a principled and highly efficient tool for diagnosing LLM behavior.</abstract>
<identifier type="citekey">ma-2026-prc</identifier>
<location>
<url>https://aclanthology.org/2026.aacl-long.10/</url>
</location>
<part>
<date>2026-11</date>
<extent unit="page">
<start>204</start>
<end>222</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T PRC: An Efficient and Unified Framework for Evaluating Robustness and Controllability of Large Language Models on Few-shot Prompt Variations
%A Ma, Chunpeng
%Y Zhang, Yue
%Y Xu, Ruifeng
%Y Purwarianti, Ayu
%Y Rong, Tong
%Y Suzuki, Jun
%S Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)
%D 2026
%8 November
%I Association for Computational Linguistics
%C Hengqin, China
%@ 979-8-89176-439-2
%F ma-2026-prc
%X Understanding the impact of prompt variations on large language model (LLM) outputs is critical. Despite extensive research, two challenges persist: conflated terminology stemming from loosely defined concepts, and inefficient evaluations reliant on output generation. To address these, we first formalize prompt categorization to clearly define prompt robustness and controllability. We then introduce the prompt vector (PV), a compact internal representation that reformulates these properties at the representation level. Building on PV, we propose PRC, an efficient and unified framework to systematically evaluate LLMs under few-shot prompt variations without requiring costly autoregressive decoding. Experiments across diverse instruction-tuned LLMs demonstrate that PRC tightly correlates with traditional output-based metrics while substantially reducing computational overhead. Overall, PRC offers a principled and highly efficient tool for diagnosing LLM behavior.
%U https://aclanthology.org/2026.aacl-long.10/
%P 204-222
Markdown (Informal)
[PRC: An Efficient and Unified Framework for Evaluating Robustness and Controllability of Large Language Models on Few-shot Prompt Variations](https://aclanthology.org/2026.aacl-long.10/) (Ma, AACL-IJCNLP 2026)
ACL