New framework HOMER uses multi-role LLMs for humor caption generation

Researchers have developed a novel framework called HOMER for generating humorous captions for images, addressing limitations in creativity and interpretability of existing large language model (LLM) approaches. This framework utilizes a humor theory called GTVH and involves a multi-role LLM collaboration. The system includes a conflicting-script extractor, a retrieval-augmented hierarchical imaginator, and a caption generator, which together aim to produce funny and diverse captions by focusing on script oppositions and creative expansion. AI

IMPACT This research introduces a novel approach to multi-modal humor generation, potentially improving LLM creativity and interpretability in creative tasks.

RANK_REASON The cluster contains an academic paper detailing a new framework for humor caption generation. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CL →

• arXiv
• GTVH
• HOMER
• Hugging Face
• New Yorker Cartoon Caption Contest Podcast
• Wenbo Shang

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →