Thai Humor Generation by Small Language Models
Published in ECTI-CON 2025, 2025
Generating humor that aligns with Thai cultural and linguistic nuances is challenging for generative AI. This paper addresses this by fine-tuning small language models (SLMs) on high-quality, non-synthetic Thai humor datasets. The Llama-3.2-3B model and Low-Rank Adaptation (LoRA) were used for efficient parameter tuning. A comprehensive evaluation with Thai participants ensured the generated humor resonated with cultural expectations.
Recommended citation: Poobanchean, P., Chaithong, S., Muansuwan, N., Limseelo, C., Sirinaovakul, B., & Suwannahong, K. (2025). Thai Humor Generation by Small Language Models. King Mongkut’s University of Technology Thonburi.
Download Paper