We categorize the world using labels that aid memory, recognition, and generalization. While some concepts have clear boundaries, others are more fluid, leading to cross-linguistic differences. How bilinguals manage these differences remains unclear. We investigate this by comparing English monolinguals, Mandarin monolinguals, and Mandarin-English bilinguals in a 2AFC task to test whether bilinguals’ categorization aligns with monolingual norms or forms an integrated system. Additionally, we develop a neural network model to simulate category boundary formation under varying language exposure. Our model closely mirrors behavioral data, supporting the idea that bilinguals develop a shared categorization system shaped by dominant language exposure. This combined behavioral and computational approach offers new insights into how bilinguals resolve cross-linguistic conflict and the cognitive mechanisms underlying multilingual concept organization.