Summary: CoCoG, the Concept based Controllable Generation framework, introduces an approach to generate visual stimuli based on controlled human concepts. This framework integrates a prediction model for human behavior in visual similarity judgments and a generation model for stimuli control.

• Enhances understanding of human cognitive processes
• Employs controllable generation to manipulate human similarity judgment
• Predicts human behavior with 64.07% accuracy using concept embeddings
• Enables experimentation with diverse object generation through concept control

The CoCoG framework is transformative as it offers tools that can influence human cognition through visual models. Its implications extend to cognitive science and AI interactions, providing deeper insights into human concept processing.