Howard University-AI4PC at SemEval-2026 Task 1: Exploring Prompt Strategies for Automatic Humor Generation

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

Summary

Howard University-AI4PC's solution system for SemEval-2026 Task 1-Subtask A, a humor generation challenge, achieved a shared 2nd place among 32 systems with an Elo score of 1020. The system, developed by Lawal Abdulmujeeb and Saurav Aryal, utilized the Llama-3.1-8B-Instruct model without any fine-tuning, demonstrating the power of prompt engineering. For news headline inputs, the team employed a two-shot prompt, framing the output as a tweet, where tone specification significantly impacted quality. When generating humor from word-pair inputs, the model was instructed to maintain a consistent everyday situation, preventing abrupt context shifts often seen when incorporating both words. The approach also incorporated specific personas, such as Dave Chappelle, to enhance humor generation. This performance highlights that competitive results in complex NLP tasks can be achieved through meticulous prompt design alone.

Key takeaway

For NLP Engineers developing creative text generation systems, you should prioritize meticulous prompt engineering before considering fine-tuning. Your focus on elements like two-shot prompting, explicit tone specification, maintaining consistent situational context, and integrating specific personas can yield highly competitive results, as demonstrated by achieving 2nd place in SemEval-2026 Task 1 without model retraining. This approach saves computational resources and development time.

Key insights

Careful prompt design with large language models can achieve competitive humor generation results without fine-tuning.

Principles

• Tone specification improves humor quality.
• Consistent context prevents abrupt shifts.
• Personas enhance humor generation.

Method

The system used Llama-3.1-8B-Instruct. Headline inputs used a two-shot tweet-framed prompt with tone. Word-pair inputs required committing to an everyday situation and leveraging personas like Dave Chappelle.

In practice

• Use two-shot prompts for specific formats.
• Specify output tone for quality.
• Employ personas for creative generation.

Topics

• Humor Generation
• Prompt Engineering
• Large Language Models
• Llama-3.1-8B-Instruct
• SemEval-2026
• Natural Language Processing
• Persona Prompting

Best for: AI Engineer, Machine Learning Engineer, Research Scientist, AI Scientist, NLP Engineer, Prompt Engineer

Related on AIssential

• 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.
• 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.
• Right Question, Right Result: Using Artificial Intelligence More Efficiently with Prompt Engineering — The quality of AI output directly correlates with the quality and specificity of the user's prompt.
• Large Language Models in NLP — LLMs integrate NLP advancements with massive scale to create general-purpose, prompt-driven language systems.
• Monthly "Is there a tool for..." Post — AI-assisted content creation requires robust prompt management and factual consistency for repeatable outputs.
• Stop Being Nice to Your Chatbot: The Surprising Science of Prompt Politeness — Polite prompts reduce LLM accuracy and increase computational cost, while direct or emotionally-charged prompts improve performance.

Open in AIssential →

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