@inproceedings{niu-zhang-2026-sici,
title = "{SICI}: A Semantic-Pragmatic Complexity Index Reveals Regime Shifts in {LLM} Stance Detection",
author = "Niu, Fuqiang and
Zhang, Bowen",
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.1836/",
doi = "10.18653/2026.findings-emnlp.1836",
pages = "39209--39221",
ISBN = "979-8-89176-462-0",
abstract = "Prompt-based LLMs are increasingly used for stance detection, but harder examples are not always repaired by clearer instructions, reasoning prompts, retrieval, or debate. We introduce (Stance Inference Complexity Index), a seven-dimensional diagnostic measure of the semantic-pragmatic burden imposed by a target{--}text pair. Across SemEval-2016 and VAST, predicts LLM accuracy better than surface proxies and shows substantial cross-scorer reliability ($\alpha=0.771$). More importantly, LLM errors change regime as increases: low-complexity examples invite over-attribution, especially \textsc{Against} predictions; intermediate examples form an unstable boundary; and high-complexity examples rapidly concentrate on \textsc{None}. This phase-transition-like structure persists across GPT-3.5, GPT-4o-mini, DeepSeek-V3, and GPT-4o, although stronger models move the boundaries. A 14-method intervention study further shows that prompting, retrieval, and debate often shift models along the attribution{--}abstention axis rather than removing the high-complexity bottleneck."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="niu-zhang-2026-sici">
<titleInfo>
<title>SICI: A Semantic-Pragmatic Complexity Index Reveals Regime Shifts in LLM Stance Detection</title>
</titleInfo>
<name type="personal">
<namePart type="given">Fuqiang</namePart>
<namePart type="family">Niu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Bowen</namePart>
<namePart type="family">Zhang</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>Prompt-based LLMs are increasingly used for stance detection, but harder examples are not always repaired by clearer instructions, reasoning prompts, retrieval, or debate. We introduce (Stance Inference Complexity Index), a seven-dimensional diagnostic measure of the semantic-pragmatic burden imposed by a target–text pair. Across SemEval-2016 and VAST, predicts LLM accuracy better than surface proxies and shows substantial cross-scorer reliability (α=0.771). More importantly, LLM errors change regime as increases: low-complexity examples invite over-attribution, especially Against predictions; intermediate examples form an unstable boundary; and high-complexity examples rapidly concentrate on None. This phase-transition-like structure persists across GPT-3.5, GPT-4o-mini, DeepSeek-V3, and GPT-4o, although stronger models move the boundaries. A 14-method intervention study further shows that prompting, retrieval, and debate often shift models along the attribution–abstention axis rather than removing the high-complexity bottleneck.</abstract>
<identifier type="citekey">niu-zhang-2026-sici</identifier>
<identifier type="doi">10.18653/2026.findings-emnlp.1836</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.1836/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>39209</start>
<end>39221</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T SICI: A Semantic-Pragmatic Complexity Index Reveals Regime Shifts in LLM Stance Detection
%A Niu, Fuqiang
%A Zhang, Bowen
%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 niu-zhang-2026-sici
%X Prompt-based LLMs are increasingly used for stance detection, but harder examples are not always repaired by clearer instructions, reasoning prompts, retrieval, or debate. We introduce (Stance Inference Complexity Index), a seven-dimensional diagnostic measure of the semantic-pragmatic burden imposed by a target–text pair. Across SemEval-2016 and VAST, predicts LLM accuracy better than surface proxies and shows substantial cross-scorer reliability (α=0.771). More importantly, LLM errors change regime as increases: low-complexity examples invite over-attribution, especially Against predictions; intermediate examples form an unstable boundary; and high-complexity examples rapidly concentrate on None. This phase-transition-like structure persists across GPT-3.5, GPT-4o-mini, DeepSeek-V3, and GPT-4o, although stronger models move the boundaries. A 14-method intervention study further shows that prompting, retrieval, and debate often shift models along the attribution–abstention axis rather than removing the high-complexity bottleneck.
%R 10.18653/2026.findings-emnlp.1836
%U https://aclanthology.org/2026.findings-emnlp.1836/
%U https://doi.org/10.18653/2026.findings-emnlp.1836
%P 39209-39221
Markdown (Informal)
[SICI: A Semantic-Pragmatic Complexity Index Reveals Regime Shifts in LLM Stance Detection](https://aclanthology.org/2026.findings-emnlp.1836/) (Niu & Zhang, Findings 2026)
ACL