Geometry of Knowledge Allows Extending Diversity Boundaries of Large Language Models

ACL ARR 2026 January Submission4967 Authors

Keywords: continuous semantic conditioning, embedding-space control of LLMs, manifold exploration, inference-time diversity, projector-based conditioning, NoveltyBench, divergent thinking

Abstract:

Starting from the hypothesis that knowledge in semantic space is organized along structured manifolds, we argue that this geometric structure renders the space explorable. By traversing it and using the resulting continuous representations to condition an LLM’s generation distribution, we can systematically expand the model’s reachable semantic range.

We introduce a framework that requires no modification of LLM parameters and operationalizes this idea by constructing a conditioning distribution from a small set of diverse anchor generations. This distribution conditions LLM's generation via an xRAG-style projector.

Our experiments demonstrate that this manifold-based conditioning substantially increases generative diversity, with direct benefits for enhancing divergent thinking, a core facet of creativity, in language models.

Paper Type: Long

Research Area: Natural Language Generation

Research Area Keywords: Natural Language Generation

Contribution Types: NLP engineering experiment, Theory

Languages Studied: English

Submission Number: 4967