Humor in AI: Massive Scale Crowd-Sourced Preferences and Benchmarks for Cartoon Captioning

TL;DR

This paper introduces a large-scale dataset of human ratings for cartoon captions, proposes new benchmarks for humor quality assessment, and evaluates current models, revealing their limitations in generating humorous content.

Contribution

It provides a massive crowd-sourced dataset, new evaluation benchmarks, and insights into the performance gaps of current AI models in humor generation.

Findings

Current models underperform top human humorists.

Fine-tuning methods like RLHF and DPO have limitations for creative tasks.

The dataset enables future research in AI humor understanding and generation.

Abstract

We present a novel multimodal preference dataset for creative tasks, consisting of over 250 million human ratings on more than 2.2 million captions, collected through crowdsourcing rating data for The New Yorker's weekly cartoon caption contest over the past eight years. This unique dataset supports the development and evaluation of multimodal large language models and preference-based fine-tuning algorithms for humorous caption generation. We propose novel benchmarks for judging the quality of model-generated captions, utilizing both GPT4 and human judgments to establish ranking-based evaluation strategies. Our experimental results highlight the limitations of current fine-tuning methods, such as RLHF and DPO, when applied to creative tasks. Furthermore, we demonstrate that even state-of-the-art models like GPT4 and Claude currently underperform top human contestants in generating humorous captions. As we conclude this extensive data collection effort, we release the entire preference dataset to the research community, fostering further advancements in AI humor generation and evaluation.