@inproceedings{chang-etal-2026-care,
title = "{CARE}: Bounded Selective Evidence-to-Judgment Routing for Cost-Aware Document Security Classification",
author = "Chang, Yeseul and
Lim, Woosang and
Kailasa, Rahul and
Shim, Simon and
Lee, Jaewoo",
editor = "Atanasova, Pepa and
Haffari, Gholamreza and
Wang, Minghan",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-industry.263/",
pages = "4423--4433",
ISBN = "979-8-89176-460-6",
abstract = "Enterprise document security classification is a selective-decision problem. Rare high-risk labels require evidence and auditability, yet judging every document with an LLM is costly and adds model-specific variance to cases that precedents already settle. We introduce CARE (Cost-Aware Resource-adaptive Evidence), a bounded, fixed commit{--}verify{--}adjudicate controller that routes each case by precedent trust. It forms a zero-LLM evidence vote, commits when a learned reliability gate predicts low vote-error risk, verifies uncertain cases through verifier{--}evidence agreement, and spends heterogeneous panel judgment only on the residual conflicts where evidence and verification disagree. Across two contrasting evidence regimes that defeat different fixed baselines, the same controller selects route depth document by document. On WikiLeaks, where contrastive retrieval already supports strong performance without LLM calls, CARE preserves the zero-call evidence endpoint within bootstrap uncertainty while staying above every always-LLM decoder. It reaches 0.9824 Macro-F1 and 0.9624 SECRET F1, with only 44 of 2,863 documents reaching any LLM stage. This requires 55 invocations in total, 1.9{\%} of a one-call-per-document budget. On DISC, where the rare TOP SECRET tail destabilizes sparse and LLM decoders, it gives the strongest row in our controlled comparison (0.8785 Macro-F1, 0.8444 TOP SECRET F1) using 496 physical model invocations against 789 for a same-evidence always-LLM policy, 63{\%} of that budget, with only 81 documents reaching the panel. CARE thus turns security classification into an auditable policy for deciding when evidence is enough and when judgment is worth spending."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="chang-etal-2026-care">
<titleInfo>
<title>CARE: Bounded Selective Evidence-to-Judgment Routing for Cost-Aware Document Security Classification</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yeseul</namePart>
<namePart type="family">Chang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Woosang</namePart>
<namePart type="family">Lim</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rahul</namePart>
<namePart type="family">Kailasa</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Simon</namePart>
<namePart type="family">Shim</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jaewoo</namePart>
<namePart type="family">Lee</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>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track</title>
</titleInfo>
<name type="personal">
<namePart type="given">Pepa</namePart>
<namePart type="family">Atanasova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Gholamreza</namePart>
<namePart type="family">Haffari</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Minghan</namePart>
<namePart type="family">Wang</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">979-8-89176-460-6</identifier>
</relatedItem>
<abstract>Enterprise document security classification is a selective-decision problem. Rare high-risk labels require evidence and auditability, yet judging every document with an LLM is costly and adds model-specific variance to cases that precedents already settle. We introduce CARE (Cost-Aware Resource-adaptive Evidence), a bounded, fixed commit–verify–adjudicate controller that routes each case by precedent trust. It forms a zero-LLM evidence vote, commits when a learned reliability gate predicts low vote-error risk, verifies uncertain cases through verifier–evidence agreement, and spends heterogeneous panel judgment only on the residual conflicts where evidence and verification disagree. Across two contrasting evidence regimes that defeat different fixed baselines, the same controller selects route depth document by document. On WikiLeaks, where contrastive retrieval already supports strong performance without LLM calls, CARE preserves the zero-call evidence endpoint within bootstrap uncertainty while staying above every always-LLM decoder. It reaches 0.9824 Macro-F1 and 0.9624 SECRET F1, with only 44 of 2,863 documents reaching any LLM stage. This requires 55 invocations in total, 1.9% of a one-call-per-document budget. On DISC, where the rare TOP SECRET tail destabilizes sparse and LLM decoders, it gives the strongest row in our controlled comparison (0.8785 Macro-F1, 0.8444 TOP SECRET F1) using 496 physical model invocations against 789 for a same-evidence always-LLM policy, 63% of that budget, with only 81 documents reaching the panel. CARE thus turns security classification into an auditable policy for deciding when evidence is enough and when judgment is worth spending.</abstract>
<identifier type="citekey">chang-etal-2026-care</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-industry.263/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>4423</start>
<end>4433</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T CARE: Bounded Selective Evidence-to-Judgment Routing for Cost-Aware Document Security Classification
%A Chang, Yeseul
%A Lim, Woosang
%A Kailasa, Rahul
%A Shim, Simon
%A Lee, Jaewoo
%Y Atanasova, Pepa
%Y Haffari, Gholamreza
%Y Wang, Minghan
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-460-6
%F chang-etal-2026-care
%X Enterprise document security classification is a selective-decision problem. Rare high-risk labels require evidence and auditability, yet judging every document with an LLM is costly and adds model-specific variance to cases that precedents already settle. We introduce CARE (Cost-Aware Resource-adaptive Evidence), a bounded, fixed commit–verify–adjudicate controller that routes each case by precedent trust. It forms a zero-LLM evidence vote, commits when a learned reliability gate predicts low vote-error risk, verifies uncertain cases through verifier–evidence agreement, and spends heterogeneous panel judgment only on the residual conflicts where evidence and verification disagree. Across two contrasting evidence regimes that defeat different fixed baselines, the same controller selects route depth document by document. On WikiLeaks, where contrastive retrieval already supports strong performance without LLM calls, CARE preserves the zero-call evidence endpoint within bootstrap uncertainty while staying above every always-LLM decoder. It reaches 0.9824 Macro-F1 and 0.9624 SECRET F1, with only 44 of 2,863 documents reaching any LLM stage. This requires 55 invocations in total, 1.9% of a one-call-per-document budget. On DISC, where the rare TOP SECRET tail destabilizes sparse and LLM decoders, it gives the strongest row in our controlled comparison (0.8785 Macro-F1, 0.8444 TOP SECRET F1) using 496 physical model invocations against 789 for a same-evidence always-LLM policy, 63% of that budget, with only 81 documents reaching the panel. CARE thus turns security classification into an auditable policy for deciding when evidence is enough and when judgment is worth spending.
%U https://aclanthology.org/2026.emnlp-industry.263/
%P 4423-4433
Markdown (Informal)
[CARE: Bounded Selective Evidence-to-Judgment Routing for Cost-Aware Document Security Classification](https://aclanthology.org/2026.emnlp-industry.263/) (Chang et al., EMNLP 2026)
ACL