Deceptive Humor: A Synthetic Multilingual Benchmark Dataset for Bridging Fabricated Claims with Humorous Content

ICLR 2026 Conference Submission399 Authors

Keywords: Deceptive Humor Detection, Multilingual Benchmark Dataset, Humor-Driven Misinformation

TL;DR: The Deceptive Humor Dataset (DHD) is a 9K-sample synthetic multilingual benchmark of humor-infused comments tied to fabricated claims, introducing a new research direction at the intersection of humor and misinformation.

Abstract: In the evolving landscape of online discourse, misinformation increasingly adopts humorous tones to evade detection and gain traction. This work introduces Deceptive Humor as a new research direction, emphasizing how false narratives, when coated in humor, become more difficult to detect and more likely to spread. To support research in this space, we present the Deceptive Humor Dataset (DHD), a multilingual collection of humor-infused comments derived from fabricated claims using the ChatGPT-4o model. Each entry is annotated with a Satire Level (from 1 for subtle satire to 3 for overt satire) and categorized into five humor types: Dark Humor, Irony, Social Commentary, Wordplay, and Absurdity. The dataset spans English, Telugu, Hindi, Kannada, Tamil, and their code-mixed forms, making it a valuable resource for multilingual analysis. Building on this foundation, we propose DH-MTL (Deceptive Humor Multi-Task Learning), a lightweight neural framework that jointly models satire intensity and humor type through a two-stage training pipeline that first adapts the encoder to deceptive humor patterns and then refines task-specific reasoning. Together, DHD and DH-MTL establish both a benchmark resource and a methodological baseline for studying how false narratives are framed, normalized, or obscured through humor.

Primary Area: datasets and benchmarks

Submission Number: 399