@inproceedings{kim-etal-2026-telling,
title = "Telling Incomplete Truths: Measuring Omission-Induced Misleadingness in Aligned Language Models",
author = "Kim, Minje and
Lee, Keeheon and
Lee, Chi-Guhn",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.656/",
pages = "15068--15089",
ISBN = "979-8-89176-459-0",
abstract = "Aligned language models are evaluated for whether they state falsehoods and whether users prefer their answers. Neither axis asks whether an answer carries what a reader needs to decide. We name that failure mode Pragmatic Reticence and build an instrument to measure it. A generator answers a passage-and-question prompt in free form, never seeing the answer options, and a frozen reader, never seeing the passage, decides from that answer alone. A natural language inference judge scores whether the answer conveys the passage{'}s decision-critical sentence. Across 47 generators the median non-reasoning model reaches a critical-fact recall of 0.28, while every reasoning-tuned checkpoint exceeds 0.74 on the whole response. Supervised fine-tuning and preference optimization do not close that gap. For models that emit a reasoning trace, the recalled fact usually sits inside the trace rather than in the answer that follows it, so a deployment that hides the trace hides the recall. In those models a trace-free prompt restores it, so the omission is a format effect. An answer can be true, preferred, and still not enough to act on."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="kim-etal-2026-telling">
<titleInfo>
<title>Telling Incomplete Truths: Measuring Omission-Induced Misleadingness in Aligned Language Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Minje</namePart>
<namePart type="family">Kim</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Keeheon</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chi-Guhn</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</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">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>Aligned language models are evaluated for whether they state falsehoods and whether users prefer their answers. Neither axis asks whether an answer carries what a reader needs to decide. We name that failure mode Pragmatic Reticence and build an instrument to measure it. A generator answers a passage-and-question prompt in free form, never seeing the answer options, and a frozen reader, never seeing the passage, decides from that answer alone. A natural language inference judge scores whether the answer conveys the passage’s decision-critical sentence. Across 47 generators the median non-reasoning model reaches a critical-fact recall of 0.28, while every reasoning-tuned checkpoint exceeds 0.74 on the whole response. Supervised fine-tuning and preference optimization do not close that gap. For models that emit a reasoning trace, the recalled fact usually sits inside the trace rather than in the answer that follows it, so a deployment that hides the trace hides the recall. In those models a trace-free prompt restores it, so the omission is a format effect. An answer can be true, preferred, and still not enough to act on.</abstract>
<identifier type="citekey">kim-etal-2026-telling</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.656/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>15068</start>
<end>15089</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Telling Incomplete Truths: Measuring Omission-Induced Misleadingness in Aligned Language Models
%A Kim, Minje
%A Lee, Keeheon
%A Lee, Chi-Guhn
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F kim-etal-2026-telling
%X Aligned language models are evaluated for whether they state falsehoods and whether users prefer their answers. Neither axis asks whether an answer carries what a reader needs to decide. We name that failure mode Pragmatic Reticence and build an instrument to measure it. A generator answers a passage-and-question prompt in free form, never seeing the answer options, and a frozen reader, never seeing the passage, decides from that answer alone. A natural language inference judge scores whether the answer conveys the passage’s decision-critical sentence. Across 47 generators the median non-reasoning model reaches a critical-fact recall of 0.28, while every reasoning-tuned checkpoint exceeds 0.74 on the whole response. Supervised fine-tuning and preference optimization do not close that gap. For models that emit a reasoning trace, the recalled fact usually sits inside the trace rather than in the answer that follows it, so a deployment that hides the trace hides the recall. In those models a trace-free prompt restores it, so the omission is a format effect. An answer can be true, preferred, and still not enough to act on.
%U https://aclanthology.org/2026.emnlp-main.656/
%P 15068-15089
Markdown (Informal)
[Telling Incomplete Truths: Measuring Omission-Induced Misleadingness in Aligned Language Models](https://aclanthology.org/2026.emnlp-main.656/) (Kim et al., EMNLP 2026)
ACL