@inproceedings{park-etal-2026-kcsat,
title = "{KCSAT}-{ML}: Probing Reasoning Models with Nationwide-Cohort Human Difficulty",
author = "Park, Sanghee and
Kim, Geewook and
Kim, Kee-Eung",
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.2468/",
pages = "53106--53129",
ISBN = "979-8-89176-462-0",
abstract = "Math reasoning benchmarks have prolifer- ated, yet most lack a per-item difficulty sig- nal grounded in actual human performance. We introduce KCSAT-ML, a decade (2014{--} 2025) of Korean College Scholastic Ability Test (KCSAT; Suneung) mathematics: 664 problems with a 339-item core set carrying official per-item error rates from nationwide cohorts of hundreds of thousands of examinees. We pair the benchmark with Difficulty-aligned Reasoning Gain (DRG): a score-orthogonal metric that asks whether a model{'}s mistakes concentrate on the items humans found hard, or on items humans found easy. Together they expose, across a wide range of VLMs (and LLMs with OCR), three patterns: (i) low- budget accuracy collapses on the high-human- error tail at every model size; (ii) test-time scaling (TTS) raises token use roughly linearly with cohort error rate, while accuracy gains follow a non-monotonic curve; (iii) within a single family, TTS flips between anti-scaling on the hardest items and overthinking on eas- ier ones {---} two faces of the same alignment failure. With the proposed DRG metric, we find that models with near-identical accuracy can sit at near-opposite values: one model gets wrong what humans also find hard, while another solves the hardest items yet fails on items humans find easy, which is a critical con- trast that aggregate accuracy hides. Our code and dataset builder is fully open-sourced at https://github.com/naver-ai/KCSAT-ML."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="park-etal-2026-kcsat">
<titleInfo>
<title>KCSAT-ML: Probing Reasoning Models with Nationwide-Cohort Human Difficulty</title>
</titleInfo>
<name type="personal">
<namePart type="given">Sanghee</namePart>
<namePart type="family">Park</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Geewook</namePart>
<namePart type="family">Kim</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kee-Eung</namePart>
<namePart type="family">Kim</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>Math reasoning benchmarks have prolifer- ated, yet most lack a per-item difficulty sig- nal grounded in actual human performance. We introduce KCSAT-ML, a decade (2014– 2025) of Korean College Scholastic Ability Test (KCSAT; Suneung) mathematics: 664 problems with a 339-item core set carrying official per-item error rates from nationwide cohorts of hundreds of thousands of examinees. We pair the benchmark with Difficulty-aligned Reasoning Gain (DRG): a score-orthogonal metric that asks whether a model’s mistakes concentrate on the items humans found hard, or on items humans found easy. Together they expose, across a wide range of VLMs (and LLMs with OCR), three patterns: (i) low- budget accuracy collapses on the high-human- error tail at every model size; (ii) test-time scaling (TTS) raises token use roughly linearly with cohort error rate, while accuracy gains follow a non-monotonic curve; (iii) within a single family, TTS flips between anti-scaling on the hardest items and overthinking on eas- ier ones — two faces of the same alignment failure. With the proposed DRG metric, we find that models with near-identical accuracy can sit at near-opposite values: one model gets wrong what humans also find hard, while another solves the hardest items yet fails on items humans find easy, which is a critical con- trast that aggregate accuracy hides. Our code and dataset builder is fully open-sourced at https://github.com/naver-ai/KCSAT-ML.</abstract>
<identifier type="citekey">park-etal-2026-kcsat</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.2468/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>53106</start>
<end>53129</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T KCSAT-ML: Probing Reasoning Models with Nationwide-Cohort Human Difficulty
%A Park, Sanghee
%A Kim, Geewook
%A Kim, Kee-Eung
%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 park-etal-2026-kcsat
%X Math reasoning benchmarks have prolifer- ated, yet most lack a per-item difficulty sig- nal grounded in actual human performance. We introduce KCSAT-ML, a decade (2014– 2025) of Korean College Scholastic Ability Test (KCSAT; Suneung) mathematics: 664 problems with a 339-item core set carrying official per-item error rates from nationwide cohorts of hundreds of thousands of examinees. We pair the benchmark with Difficulty-aligned Reasoning Gain (DRG): a score-orthogonal metric that asks whether a model’s mistakes concentrate on the items humans found hard, or on items humans found easy. Together they expose, across a wide range of VLMs (and LLMs with OCR), three patterns: (i) low- budget accuracy collapses on the high-human- error tail at every model size; (ii) test-time scaling (TTS) raises token use roughly linearly with cohort error rate, while accuracy gains follow a non-monotonic curve; (iii) within a single family, TTS flips between anti-scaling on the hardest items and overthinking on eas- ier ones — two faces of the same alignment failure. With the proposed DRG metric, we find that models with near-identical accuracy can sit at near-opposite values: one model gets wrong what humans also find hard, while another solves the hardest items yet fails on items humans find easy, which is a critical con- trast that aggregate accuracy hides. Our code and dataset builder is fully open-sourced at https://github.com/naver-ai/KCSAT-ML.
%U https://aclanthology.org/2026.findings-emnlp.2468/
%P 53106-53129
Markdown (Informal)
[KCSAT-ML: Probing Reasoning Models with Nationwide-Cohort Human Difficulty](https://aclanthology.org/2026.findings-emnlp.2468/) (Park et al., Findings 2026)
ACL