CreativityNeuro: Steering Language Model Weights to Improve Divergent Thinking and Reduce Mode Collapse

2026-07-01

AI Summary

Large language models often produce convergent responses in open-ended tasks, exhibiting “artificial hive mind” behavior and suffering from mode collapse. This work proposes CreativityNeuro, a training-free, gradient-free, and behavioral-data-free method that enhances divergent thinking by directly modulating model weights in the weight space through contrastive weight manipulation. It represents the first approach to achieve weight-level creativity steering without fine-tuning, substantially improving generalization. Experimental results demonstrate that CreativityNeuro boosts performance by up to 14 human percentile points on the Divergent Association Task (DAT) and significantly enhances response originality, surprise, and overall creativity in human evaluations on the Alternate Uses Task (AUT) and Task Task, while effectively mitigating mode collapse.

Abstract

Divergent thinking is a crucial aspect of creativity, yet large language models (LLMs) tend to consistently generate similar responses to open-ended questions, in what has been termed the artificial hivemind effect. Here, we introduce CreativityNeuro, a data-free method for enhancing divergent thinking in LLMs via contrastive weight steering. We evaluate our method across multiple creativity assessments and report several main findings. On the Divergent Association Task (DAT), a vocabulary-space creativity test, CreativityNeuro improves performance by up to 14 human percentile points. Next, in a large-scale human evaluation (N=720) on the Alternative Uses Test (AUT) and the Task Task, CreativityNeuro achieves significant improvements in originality, surprise, and creativity, transferring to longer-form and more open-ended tasks. Importantly, we find that across all three tasks, CreativityNeuro demonstrably reduces measures of mode collapse. Moreover, activation steering achieves comparable performance to CreativityNeuro on the DAT, but it does not transfer to the AUT and Task Task, demonstrating the effectiveness of weight-space steering in generalizing to unseen tasks. In conclusion, CreativityNeuro improves divergent thinking and reduces mode collapse without requiring behavioral data, re-training, or gradient-based fine-tuning, providing a straightforward way to enhance LLM performance in creative domains.

Problem

• divergent thinking
• mode collapse
• large language models
• creativity
• artificial hivemind

Innovation

• weight steering
• divergent thinking
• mode collapse
• data-free method
• creativity enhancement