Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
The paper introduces CREATE, a benchmark to evaluate large language models' (LLMs) capacity for creative associative reasoning by generating diverse and specific paths connecting concepts within their parametric knowledge. The benchmark scores models based on the number, specificity, and diversity of the generated paths. Experiments on frontier models reveal that stronger models achieve higher creative utility, but the complexity of the search space makes benchmark saturation difficult, and "thinking" models don't always perform better.
LLMs struggle to generate diverse and specific connections between concepts, even with high token budgets and "thinking" prompts, revealing a gap in creative associative reasoning.
A key component of creativity is associative reasoning: the ability to draw novel yet meaningful connections between concepts. We introduce CREATE, a benchmark designed to evaluate models'capacity for creative associative reasoning. CREATE requires models to generate sets of paths connecting concepts in a model's parametric knowledge. Paths should have high specificity (distinctiveness and closeness of the concept connection) and high diversity (dissimilarity from other paths), and models are scored more highly if they produce a larger set of strong, diverse paths. This task shares demands of real creativity tasks like hypothesis generation, including an extremely large search space, but enables collection of a sizable benchmark with objective answer grading. Evaluation of frontier models shows that the strongest models achieve higher creative utility than others, with the high multiplicity of answers and complexity of the search making benchmark saturation difficult to achieve. Furthermore, our results illustrate that thinking models are not always more effective on our task, even with high token budgets. Recent approaches for creative prompting give some but limited additional improvement. CREATE provides a sandbox for developing new methods to improve models'capacity for associative creativity.
Meta AI