Collaborative virtual environment (CVE) systems support many users to share a virtual world and to do collaborative work and entities exchange information on network by sending messages to maintain the consistency of the systems. To maintain the scalability, interest management is usually adopted in designing a large-scale CVE system. Entities in the CVE system are divided into groups and each entity only sends updates to the entities that are interested in it in order to reduce network traffic. This paper describes several grouping strategies: entity-based grouping and cellbased grouping and proposes an adaptive grouping scheme based on template matching for the tradeoff of reducing the superfluous messages and the cost of maintaining groups in order to minimize the overhead of grouping. The simulation results show that the scheme is fit for the implementation of a cost-effective and high performance CVE system.