Elamur Research⌕
☰
home / papers / 2026 / Humor Generation – Text-based Humor Generation (English)

Humor Generation – Text-based Humor Generation (English)

Paper · 2026 · Hemeshkumar Parthiban and Priyadharsini R

Summary

The paper describes a system for SemEval-2026 Task 1 (MWAHAHA), Subtask A. In this English humor-generation task, a model gets either a news headline or a pair of unusual words that must both appear, and it must produce an original, safe joke. The authors use no training. Their pipeline has four steps: light input cleanup, a single fixed instruction template, zero-shot inference with a GPT-4–class model under deterministic decoding, and an automatic check that regenerates outputs that break the format rules. The template asks for one sentence of 8–12 words in a deadpan, understated style that gently undercuts expectations and avoids loud punchlines or unsafe content.

On the human-judged leaderboard, the system finished in the middle of the field. A qualitative ablation suggests that fuller constraints give more stylistically fitting output. The work shows that prompt-only control is a cheap, reproducible baseline for constrained creative generation. However, the evidence for quality improvements is limited.

Key points

  • Task inputs: Subtask A has two settings, Word Inclusion (two rare words that must both be used) and News Headline (a satirical sentence based on a real headline). Outputs are ranked by human comparative judgment.
  • Leaderboard result: The submission "hemeshkumar_31" scored 843 (95% CI [802, 875]) and placed 31st. Top systems scored above 1000.
  • Relative performance claim: The authors say this is about 84% of the top score, but that figure assumes a top score of roughly 1000.
  • Prompt ablation: Three prompts were compared (unconstrained, tone-only, full constraints). The comparison was qualitative only, with no numeric scores reported.
  • Error analysis: Outputs were sometimes well-formed but not funny, generic in their satire, overly literal, or awkwardly compressed by the length limit. Suggested future work includes adaptive prompts, fine-tuning, and incongruity modeling.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

Also on the byline: Rajalakshmi Engineering College.

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 worksPsychology · 1,649 worksSocial Psychology · 330 works

🧭 Topics 1Research subjects and languages the work is about.English · 2,379 works

Cite

Cite as Hemeshkumar Parthiban and Priyadharsini R. 2026. Humor Generation – Text-based Humor Generation (English). Proceedings of the 20th International Workshop on Semantic Evaluation (2026).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.semeval-1.104

@inproceedings{parthiban2026humor,
  title = {{Humor Generation – Text-based Humor Generation (English)}},
  author = {Hemeshkumar Parthiban and Priyadharsini R},
  booktitle = {Proceedings of the 20th International Workshop on Semantic Evaluation (2026)},
  year = {2026},
  doi = {10.18653/v1/2026.semeval-1.104},
  url = {https://doi.org/10.18653/v1/2026.semeval-1.104}
}

References

8 references, 3 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Tom Brown and 1 others. 2020. Language models are few-shot learners. Advances in Neural Information Processing Systems.
  2. Md Kamrul Hasan and 1 others. 2019. Ur-funny: A multimodal language dataset for understanding humor. In Proceedings of EMNLP. find
  3. Pengfei Liu and 1 others. 2023. Pre-train, prompt, and predict: A systematic survey of prompting methods. ACM Computing Surveys. find
  4. Xiaotong Liu, Carlos Fernandez, and Rohan Mehta. 2025. Constraint-guided generation in large language models. In Proceedings of the 2025 Annual Meeting of the Association for Computational Linguistics. find
  5. Rada Mihalcea and Carlo Strapparava. 2005. Making computers laugh: Investigations in automatic humor recognition. In Proceedings of Human Language Technology Conference.
  6. Task Organizers. 2026. Semeval-2026 task 1: Mwahaha – competition on humor generation. In Proceedings of the 20th International Workshop on Semantic Evaluation. Placeholder citation; official BibTeX to be released. find
  7. Jun Wang, Aarav Patel, and Priya Singh. 2024. Evaluating large language models for creative text generation. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing. find
  8. Rui Zhang, Yifan Li, and Ming Chen. 2024. Controlled text generation with large language models: A survey. Transactions of the Association for Computational Linguistics.

Updated

This page in Markdown → https://md.research.elamur.ai/free/en/papers/parthiban-2026-humor-generation-text-based-humor/