abateam at SemEval-2026 Task 1: Plan2joke – Humor Policies for Type-Specific Two-Pass Humor Generation

2026-06-25 · Source: Paper Index on ACL Anthology · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Natural Language Processing · Depth: Expert, quick

Summary

The abateam developed Plan2joke, a policy-driven humor generation system for SemEval-2026 Task 1, focusing on type-specific two-pass humor generation. This approach integrates optimal humor recognition systems and a context enrichment strategy to cover multiple humor types. The system's design was influenced by recent computational humor research, including works by Baranov et al. (2023), Tikhonov and Shtykovskiy (2024), and Zhong et al. (2023). Plan2joke employs Supervised Fine-Tuning (SFT) on a custom dataset, which combines previous research samples and is adjusted to align with the defined humor policies. An ablation study was conducted to calibrate the system and evaluate its components.

Key takeaway

For NLP Engineers developing humor generation systems, this research suggests adopting a policy-driven, two-pass approach. You should consider integrating explicit humor policies and robust humor recognition systems to achieve type-specific humor. Fine-tuning your models with context-enriched, policy-aligned datasets can significantly improve generation quality and control. This method offers a structured way to enhance the creativity and relevance of generated jokes.

Key insights

A policy-driven, two-pass approach combining humor recognition and context enrichment can generate type-specific humor.

Principles

• Humor generation benefits from explicit policy guidance.
• Context enrichment enhances humor generation quality.
• Supervised fine-tuning is effective for policy alignment.

Method

The Plan2joke method involves designing humor policies, using optimal humor recognition and context enrichment, and applying SFT on a policy-aligned dataset, followed by ablation studies for calibration.

In practice

• Implement policy-driven humor generation.
• Integrate humor recognition systems.
• Fine-tune models with policy-aligned data.

Topics

• Computational Humor
• Humor Generation
• SemEval-2026 Task 1
• Supervised Fine-Tuning
• Natural Language Generation
• Humor Recognition

Best for: Research Scientist, AI Scientist, NLP Engineer

Related on AIssential

• Humor Style Drives Laughter, Topic Shapes Acceptability: Evaluating Bilingual Personal and Political Robot-Delivered AI Jokes — Aggressive and Affiliative humor styles enhance robot-delivered joke funniness, while person-related topics improve appropriateness.
• Automatic Generation of Highlights for Academic Paper Via Prompt-based Learning — Prompt-based learning with LLMs can generate academic paper highlights effectively without extensive labeled data.
• CAF-Gen: A Multi-Agent System for Enriching Argumentation Structures — CAF-Gen's multi-agent Creator-Reviewer system enriches shallow argument structures into robust, CAF-compliant models, overcoming single-pass generation instability.
• FOL2NS: Generating Natural Sentences from First-Order Logic — FOL2NS is a neurosymbolic framework translating complex first-order logic into natural language, addressing data scarcity.
• Token-Level Pun Location Using Multi-Layer BERT with Mixture of Experts — A two-stage BERT-MoE model effectively locates puns in Portuguese by specializing expert components.
• Re-defining Humor Data Objects for AI Humor Research — AI humor research improves by defining humor as a social interaction with context and explanations, not just a binary state.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Paper Index on ACL Anthology.