lmfaoooo at SemEval-2026 Task 1: Humor Is an Audience. Preference Modeling for Constrained Humor Generation
arXiv:2606.00022v1 Announce Type: new Abstract: Humor generation remains difficult not only because producing fluent, novel jokes is hard, but because "funny" is audience-dependent and supervision is noisy -- preferences vary with audience, context, and culture, and annotator agreement is often low. In this paper, we describe our system for the SemEval-2026 Task-1 (MWAHAHA), which focuses on humor generation under explicit constraints. The task evaluates submitted systems via human preference judgments in 1-on-1 arena-style comparisons. We adopt a "generate-many -> select-best" strategy. First
The proliferation of advanced language models makes nuanced human-like generation, such as humor, a current frontier in AI research, pushing the boundaries of AI's ability to understand and produce complex human concepts.
This work highlights the increasing sophistication and human-centric challenges in AI development, particularly around subjective concepts like humor, impacting areas from content creation to human-robot interaction.
The explicit focus on audience-dependent preference modeling for humor generation signals a shift towards more adaptive and context-aware AI systems, moving beyond purely objective performance metrics.
- · AI research community
- · Generative AI developers
- · Entertainment sector
- · Personalized content platforms
- · AI systems lacking preference modeling
- · Generic content generators
AI systems will become more adept at generating content that aligns with specific user preferences and cultural contexts.
The development of sophisticated preference models for humor could lead to more engaging and personalized interactions with AI agents.
This could accelerate the creation of truly adaptive AI companions and content engines that deeply understand and cater to individual human subjectivity.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.CL