SLPGFJWUInsa at SemEval-2026 Task 1: Enhancing Linguistic Creativity for English Text-Based Humor

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

Summary

The paper "SLPGFJWUInsa at SemEval-2026 Task 1" by Insa Abbas and Sadaf Abdul Rauf describes a joke generation system developed for SemEval-2026 Task 1, Subtask A. This system aims to create humor under constrained conditions, utilizing unusual words and news headlines as input. The model was trained on a unique dataset combining LLM-generated and human-curated augmented data, specifically designed to produce constrained humor and bridge the gap between these data types. The authors demonstrate that applying parameter-efficient fine-tuning (PEFT) on high-quality pre-trained base models, coupled with a well-crafted prompt design, enables their model to generate high-quality, innovative output while maintaining the desired stylistic constraints. This work was presented at the 20th International Workshop on Semantic Evaluation (2026) in San Diego, California, USA, with proceedings published on pages 1832–1839.

Key takeaway

For NLP Engineers developing constrained text generation systems, this research highlights the effectiveness of combining PEFT with high-quality pre-trained models. You should integrate LLM-generated and human-curated data for robust training and prioritize meticulous prompt design. This approach allows you to achieve innovative, style-consistent outputs, particularly when working with unusual inputs like news headlines for humor.

Key insights

PEFT on pre-trained LLMs with prompt design generates constrained, high-quality humor from unusual words and headlines.

Principles

• Combine LLM-generated and human-curated data.
• Leverage PEFT for efficient fine-tuning.
• Prompt design is crucial for style control.

Method

Train a model using PEFT on high-quality pre-trained base models. Use a dataset of LLM-generated and human-curated augmented data. Employ well-crafted prompt design for constrained humor generation from unusual words and news headlines.

In practice

• Generate humor from unusual words.
• Create jokes from news headlines.
• Augment data with LLM-generated content.

Topics

• Humor Generation
• Text-Based Humor
• SemEval-2026
• PEFT
• Large Language Models
• Prompt Engineering

Best for: Research Scientist, AI Scientist, NLP Engineer

Related on AIssential

• 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.
• Efficient Fine-Tuning Methods for Portuguese Question Answering: A Comparative Study of PEFT on BERTimbau and Exploratory Evaluation of Generative LLMs — PEFT methods efficiently fine-tune encoder-based models for low-resource languages, outperforming generative LLMs in resource usage.
• Large Language Models: From Prototype to Production — A pragmatic approach enables successful NLP projects using LLMs from prototype to production.
• Children's English Reading Story Generation via Supervised Fine-Tuning of Compact LLMs with Controllable Difficulty and Safety — Fine-tuning compact 8B LLMs can generate children's stories with controllable difficulty and safety, outperforming larger models.
• SciText2Eq: Assessing LLMs for Explainable Equation Generation for Scientific Creativity — LLMs moderately generate equations from text but struggle with semantic accuracy and reliable self-evaluation.
• How Can We Synthesize High-Quality Pretraining Data? A Systematic Study of Prompt Design, Generator Model, and Source Data — Structured pedagogical formats and diverse outputs are key to high-quality synthetic pretraining data, not larger generator models.

Open in AIssential →

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