English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Chain-of-Thought (CoT) guides large language models (LLMs) to reasonstep-by-step, and can motivate their logical reasoning ability. While effectivefor logical tasks, CoT is not conducive to creative problem-solving which oftenrequires out-of-box thoughts and is crucial for innovation advancements. Inthis paper, we explore the Leap-of-Thought (LoT) abilities within LLMs -- anon-sequential, creative paradigm involving strong associations and knowledgeleaps. To this end, we study LLMs on the popular Oogiri game which needsparticipants to have good creativity and strong associative thinking forresponding unexpectedly and humorously to the given image, text, or both, andthus is suitable for LoT study. Then to investigate LLMs' LoT ability in theOogiri game, we first build a multimodal and multilingual Oogiri-GO datasetwhich contains over 130,000 samples from the Oogiri game, and observe theinsufficient LoT ability or failures of most existing LLMs on the Oogiri game.Accordingly, we introduce a creative Leap-of-Thought (CLoT) paradigm to improveLLM's LoT ability. CLoT first formulates the Oogiri-GO dataset intoLoT-oriented instruction tuning data to train pretrained LLM for achievingcertain LoT humor generation and discrimination abilities. Then CLoT designs anexplorative self-refinement that encourages the LLM to generate more creativeLoT data via exploring parallels between seemingly unrelated concepts andselects high-quality data to train itself for self-refinement. CLoT not onlyexcels in humor generation in the Oogiri game but also boosts creativeabilities in various tasks like cloud guessing game and divergent associationtask. These findings advance our understanding and offer a pathway to improveLLMs' creative capacities for innovative applications across domains. Thedataset, code, and models will be released online.https://zhongshsh.github.io/CLoT/.

Let's Think Outside the Box: Exploring Leap-of-Thought in Large Language Models with Creative Humor Generation