Abstract
Innovative idea generation remains a core challenge in AI, as large languagemodels (LLMs) often struggle to produce outputs that are both novel andrelevant. Despite their fluency, LLMs tend to replicate patterns seen duringtraining, limiting their ability to diverge creatively without extensive promptengineering. Prior work has addressed this through domain-specific heuristicsand structured prompting pipelines, but such solutions are brittle anddifficult to generalize. In this paper, we propose a model-agnosticlatent-space ideation framework that enables controlled, scalable creativity bynavigating the continuous embedding space of ideas. Unlike prior methods, ourframework requires no handcrafted rules and adapts easily to different domains,input formats, and creative tasks. This paper introduces an early-stageprototype of our method, outlining the conceptual framework and preliminaryresults highlighting its potential as a general-purpose co-ideator for human-AIcollaboration.