Universe of Thoughts: Enabling Creative Reasoning with Large Language Models
Reasoning based on Large Language Models (LLMs) has garnered increasing attention due to outstanding performance of these models in mathematical and complex logical tasks. Beginning with the Chain-of-Thought (CoT) prompting technique, numerous reasoning methods have emerged that decompose problems into smaller, sequential steps (or thoughts). However, existing reasoning models focus on conventional problem-solving and do not necessarily generate creative solutions by “creative reasoning”. In domains where the solution space is expansive and conventional solutions are suboptimal, such as drug discovery or business strategization, creative reasoning to discover innovative solutions is crucial. To address this gap, first we introduce a computational framework for creative reasoning inspired by established cognitive science principles. With this framework, we propose three core creative reasoning paradigms, namely, combinational, exploratory, and transformative reasoning, where each offers specific directions for systematic exploration of the universe of thoughts to generate creative solutions. Next, to materialize this framework using LLMs, we introduce the Universe of Thoughts (or UoT, for short), a novel set of methods to implement the aforementioned three creative processes. Finally, we introduce three novel tasks that necessitate creative problem-solving, along with an evaluation benchmark to assess creativity from three orthogonal perspectives: feasibility as constraint, and utility and novelty as metrics. With a comparative analysis against the state-of-the-art (SOTA) reasoning techniques as well as representative commercial models with reasoning capability, we show that UoT demonstrates superior performance in creative reasoning. This work introduces a new perspective on how LLMs can become autonomously creative, advancing the field to address problems that require more innovative solutions.
Combinational Creative Reasoning: In this paradigm, one generates novel and creative solutions by (1) identifying known thoughts (i.e., thoughts that have been used as part of some solution in the universe of thoughts) that are relevant to the target solution space of the problem on hand but have not been previously used as part of any solution in this space, and (2) combines these thoughts with existing thoughts in the target solution space. A classic example of combinational creativity is a collage, which combines different types of existing visuals in an unconventional way to create a new piece of art.
Exploratory Creative Reasoning: This paradigm is similar to the previous paradigm, where the main difference is that the adopted thought(s) from outside of the target solution space are individual thoughts that are not necessarily part of any known solution in the universe of thoughts. In this paradigm, the new thoughts expand the target solution space by introducing new conceptual building blocks that combine with existing thoughts to generate novel solutions. Continuing with our example of painting, when the Impressionism painting style was introduced, creative artists like Monet utilized brushstrokes in a new functional way, a thought that had not been considered as part of the existing “solutions” to create a painting at that time.
Transformational Creative Reasoning: This is the most advanced paradigm of creative reasoning, as it involves fundamentally altering or dropping the core rules that define the presumed solution space and provides the opportunity to consider other solution spaces to identify creative solutions for the problem on hand. With this approach the solution space is changed, allowing for discovery of solutions that were previously inconceivable for the target problem. The example here is the Cubism painting style pioneered by Picasso, who broke from the traditional rule of direct representation by depicting objects from multiple angles simultaneously and by fragmenting forms into geometric shapes.