Thunder-KoNUBench: A Corpus-Aligned Benchmark for Korean Negation Understanding

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Jung, Sungmok, So, Yeonkyoung, Lee, Joonhak, Kim, Sangho, Ahn, Yelim, Lee, Jaejin
Format: Preprint
Publié: 2026
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866914240157188096
author Jung, Sungmok
So, Yeonkyoung
Lee, Joonhak
Kim, Sangho
Ahn, Yelim
Lee, Jaejin
author_facet Jung, Sungmok
So, Yeonkyoung
Lee, Joonhak
Kim, Sangho
Ahn, Yelim
Lee, Jaejin
contents Although negation is known to challenge large language models (LLMs), benchmarks for evaluating negation understanding, especially in Korean, are scarce. We conduct a corpus-based analysis of Korean negation and show that LLM performance degrades under negation. We then introduce Thunder-KoNUBench, a sentence-level benchmark that reflects the empirical distribution of Korean negation phenomena. Evaluating 47 LLMs, we analyze the effects of model size and instruction tuning, and show that fine-tuning on Thunder-KoNUBench improves negation understanding and broader contextual comprehension in Korean.
format Preprint
id arxiv_https___arxiv_org_abs_2601_04693
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Thunder-KoNUBench: A Corpus-Aligned Benchmark for Korean Negation Understanding
Jung, Sungmok
So, Yeonkyoung
Lee, Joonhak
Kim, Sangho
Ahn, Yelim
Lee, Jaejin
Computation and Language
Although negation is known to challenge large language models (LLMs), benchmarks for evaluating negation understanding, especially in Korean, are scarce. We conduct a corpus-based analysis of Korean negation and show that LLM performance degrades under negation. We then introduce Thunder-KoNUBench, a sentence-level benchmark that reflects the empirical distribution of Korean negation phenomena. Evaluating 47 LLMs, we analyze the effects of model size and instruction tuning, and show that fine-tuning on Thunder-KoNUBench improves negation understanding and broader contextual comprehension in Korean.
title Thunder-KoNUBench: A Corpus-Aligned Benchmark for Korean Negation Understanding
topic Computation and Language
url https://arxiv.org/abs/2601.04693