@inproceedings{lanzeray-etal-2026-mortalmath,
title = "{M}ortal{MATH}: Evaluating the Conflict Between Reasoning Objectives and Emergency Contexts",
author = "Lanzeray, Etienne and
Meilliez, Stephane and
Ruelle, Malo and
Sileo, Damien",
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 2: Short Papers)",
month = nov,
year = "2026",
address = "Hengqin, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.aacl-short.8/",
pages = "88--94",
ISBN = "979-8-89176-440-8",
abstract = "Large Language Models are increasingly optimized for deep reasoning, often through post-training objectives that strongly reward correct task completion. We ask whether this pressure toward correct completion can interfere with recognizing when a task should no longer be pursued. We introduce \textit{MortalMATH}, a benchmark of 150 scenarios in which users request algebra help while describing increasingly urgent situations, from mild distraction to life-threatening emergencies. We find a striking contrast across models: several reasoning-oriented models, including Qwen3-32B and GPT-5 variants, continue solving the mathematical task with high accuracy even under extreme emergency contexts, while other models increasingly interrupt the task or acknowledge the danger. Reasoning models also spend substantial computation on the original problem, delaying safety-relevant information by several seconds. These results expose a failure mode of highly optimized task completion: training models to persist until they reach a verifiably correct answer may also make them less willing to stop when the surrounding context demands it."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="lanzeray-etal-2026-mortalmath">
<titleInfo>
<title>MortalMATH: Evaluating the Conflict Between Reasoning Objectives and Emergency Contexts</title>
</titleInfo>
<name type="personal">
<namePart type="given">Etienne</namePart>
<namePart type="family">Lanzeray</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Stephane</namePart>
<namePart type="family">Meilliez</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Malo</namePart>
<namePart type="family">Ruelle</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Damien</namePart>
<namePart type="family">Sileo</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 2: Short 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">979-8-89176-440-8</identifier>
</relatedItem>
<abstract>Large Language Models are increasingly optimized for deep reasoning, often through post-training objectives that strongly reward correct task completion. We ask whether this pressure toward correct completion can interfere with recognizing when a task should no longer be pursued. We introduce MortalMATH, a benchmark of 150 scenarios in which users request algebra help while describing increasingly urgent situations, from mild distraction to life-threatening emergencies. We find a striking contrast across models: several reasoning-oriented models, including Qwen3-32B and GPT-5 variants, continue solving the mathematical task with high accuracy even under extreme emergency contexts, while other models increasingly interrupt the task or acknowledge the danger. Reasoning models also spend substantial computation on the original problem, delaying safety-relevant information by several seconds. These results expose a failure mode of highly optimized task completion: training models to persist until they reach a verifiably correct answer may also make them less willing to stop when the surrounding context demands it.</abstract>
<identifier type="citekey">lanzeray-etal-2026-mortalmath</identifier>
<location>
<url>https://aclanthology.org/2026.aacl-short.8/</url>
</location>
<part>
<date>2026-11</date>
<extent unit="page">
<start>88</start>
<end>94</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T MortalMATH: Evaluating the Conflict Between Reasoning Objectives and Emergency Contexts
%A Lanzeray, Etienne
%A Meilliez, Stephane
%A Ruelle, Malo
%A Sileo, Damien
%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 2: Short Papers)
%D 2026
%8 November
%I Association for Computational Linguistics
%C Hengqin, China
%@ 979-8-89176-440-8
%F lanzeray-etal-2026-mortalmath
%X Large Language Models are increasingly optimized for deep reasoning, often through post-training objectives that strongly reward correct task completion. We ask whether this pressure toward correct completion can interfere with recognizing when a task should no longer be pursued. We introduce MortalMATH, a benchmark of 150 scenarios in which users request algebra help while describing increasingly urgent situations, from mild distraction to life-threatening emergencies. We find a striking contrast across models: several reasoning-oriented models, including Qwen3-32B and GPT-5 variants, continue solving the mathematical task with high accuracy even under extreme emergency contexts, while other models increasingly interrupt the task or acknowledge the danger. Reasoning models also spend substantial computation on the original problem, delaying safety-relevant information by several seconds. These results expose a failure mode of highly optimized task completion: training models to persist until they reach a verifiably correct answer may also make them less willing to stop when the surrounding context demands it.
%U https://aclanthology.org/2026.aacl-short.8/
%P 88-94
Markdown (Informal)
[MortalMATH: Evaluating the Conflict Between Reasoning Objectives and Emergency Contexts](https://aclanthology.org/2026.aacl-short.8/) (Lanzeray et al., AACL-IJCNLP 2026)
ACL
- Etienne Lanzeray, Stephane Meilliez, Malo Ruelle, and Damien Sileo. 2026. MortalMATH: Evaluating the Conflict Between Reasoning Objectives and Emergency Contexts. In 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 2: Short Papers), pages 88–94, Hengqin, China. Association for Computational Linguistics.