YNU-HPCC at SemEval-2026 Task 1: Constraint-Aware In-Context Learning for Multilingual 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 YNU-HPCC team developed a system for SemEval-2026 Task 1, focusing on multilingual humor generation from news headlines or two unrelated words. The primary challenge addressed was enabling Large Language Models (LLMs) to comprehend human humor and adhere to specific humorous styles. The team explored two main strategies: fine-tuning LLMs using Proximal Policy Optimization (PPO) and employing in-context learning with LLMs. For PPO, a hybrid reward model was constructed to align with humor. Their final submission relied on multiple advanced LLMs, guided by customized few-shot prompts and a small set of gold samples, to generate jokes that resonated with human humor. The quality of generated texts was evaluated using Qwen-Max. This system achieved competitive results, ranking 4th in the English track, 2nd in the Chinese track, and 2nd in the Spanish track.

Key takeaway

For NLP Engineers developing multilingual humor generation systems, this work suggests prioritizing in-context learning over PPO fine-tuning for competitive results. You should focus on crafting customized few-shot prompts and curating small sets of gold samples to effectively guide Large Language Models. Additionally, consider integrating models like Qwen-Max for robust evaluation of generated humorous texts across languages. This approach can significantly improve humor alignment and task performance.

Key insights

Constraint-aware in-context learning with few-shot prompts effectively guides LLMs for competitive multilingual humor generation.

Principles

• LLMs require specific guidance for human humor.
• Hybrid reward models align PPO with humor.
• Few-shot prompts enhance LLM humor generation.

Method

The system used in-context learning with advanced LLMs, customized few-shot prompts, and gold samples for humor generation, evaluated by Qwen-Max. PPO fine-tuning with a hybrid reward model was also explored.

In practice

• Apply few-shot prompts for humor tasks.
• Evaluate humor generation with Qwen-Max.
• Investigate hybrid reward models for PPO.

Topics

• Humor Generation
• Large Language Models
• In-Context Learning
• Few-Shot Prompting
• Multilingual NLP
• SemEval-2026

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

Related on AIssential

• Supervision versus Demonstration-Based In-Context Learning for Multiword Expression Classification — In-context learning for Turkish LVCs is highly prompt-sensitive, with few-shot demonstrations improving LLM calibration to match supervised baselines.
• Linguistic Symptoms: Augmented Generation by Retrieval and Reasoning in LLMs under Portuguese-English Variation in Medical Contexts — Linguistic variation significantly impacts smaller LLMs in medical contexts, with larger models showing greater cross-language robustness.
• Loong: A Human-Like Long Document Translation Agent with Observe-and-Act Adaptive Context Selection — Loong is a human-like agent that uses adaptive context selection and reinforcement learning for long document translation.
• 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.
• Unveiling the Limits of Large Language Models in Inferring Pragmatic Meaning from Non-Verbal Responses — LLMs significantly struggle to infer pragmatic meaning from non-verbal communication, unlike their performance with verbal cues.
• 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.

Open in AIssential →

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