Oogiri-Master: Benchmarking Humor Understanding via Oogiri

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Murakami, Soichiro, Kamigaito, Hidetaka, Takamura, Hiroya, Okumura, Manabu
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911338694967296
author Murakami, Soichiro
Kamigaito, Hidetaka
Takamura, Hiroya
Okumura, Manabu
author_facet Murakami, Soichiro
Kamigaito, Hidetaka
Takamura, Hiroya
Okumura, Manabu
contents Humor is a salient testbed for human-like creative thinking in large language models (LLMs). We study humor using the Japanese creative response game Oogiri, in which participants produce witty responses to a given prompt, and ask the following research question: What makes such responses funny to humans? Previous work has offered only limited reliable means to answer this question. Existing datasets contain few candidate responses per prompt, expose popularity signals during ratings, and lack objective and comparable metrics for funniness. Thus, we introduce Oogiri-Master and Oogiri-Corpus, which are a benchmark and dataset designed to enable rigorous evaluation of humor understanding in LLMs. Each prompt is paired with approximately 100 diverse candidate responses, and funniness is rated independently by approximately 100 human judges without access to others' ratings, reducing popularity bias and enabling robust aggregation. Using Oogiri-Corpus, we conduct a quantitative analysis of the linguistic factors associated with funniness, such as text length, ambiguity, and incongruity resolution, and derive objective metrics for predicting human judgments. Subsequently, we benchmark a range of LLMs and human baselines in Oogiri-Master, demonstrating that state-of-the-art models approach human performance and that insight-augmented prompting improves the model performance. Our results provide a principled basis for evaluating and advancing humor understanding in LLMs.
format Preprint
id arxiv_https___arxiv_org_abs_2512_21494
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Oogiri-Master: Benchmarking Humor Understanding via Oogiri
Murakami, Soichiro
Kamigaito, Hidetaka
Takamura, Hiroya
Okumura, Manabu
Computation and Language
Artificial Intelligence
Humor is a salient testbed for human-like creative thinking in large language models (LLMs). We study humor using the Japanese creative response game Oogiri, in which participants produce witty responses to a given prompt, and ask the following research question: What makes such responses funny to humans? Previous work has offered only limited reliable means to answer this question. Existing datasets contain few candidate responses per prompt, expose popularity signals during ratings, and lack objective and comparable metrics for funniness. Thus, we introduce Oogiri-Master and Oogiri-Corpus, which are a benchmark and dataset designed to enable rigorous evaluation of humor understanding in LLMs. Each prompt is paired with approximately 100 diverse candidate responses, and funniness is rated independently by approximately 100 human judges without access to others' ratings, reducing popularity bias and enabling robust aggregation. Using Oogiri-Corpus, we conduct a quantitative analysis of the linguistic factors associated with funniness, such as text length, ambiguity, and incongruity resolution, and derive objective metrics for predicting human judgments. Subsequently, we benchmark a range of LLMs and human baselines in Oogiri-Master, demonstrating that state-of-the-art models approach human performance and that insight-augmented prompting improves the model performance. Our results provide a principled basis for evaluating and advancing humor understanding in LLMs.
title Oogiri-Master: Benchmarking Humor Understanding via Oogiri
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2512.21494