Core Viewpoint - The article discusses the advancements in AI 3D generation technology, highlighting the introduction of a new evaluation system called Hi3DEval, which aims to provide a comprehensive assessment of 3D content generation quality through a multi-layered approach [3][8][28]. Group 1: Hi3DEval Evaluation System - Hi3DEval is a new hierarchical automatic evaluation system for 3D content generation, developed in collaboration with several universities, addressing the limitations of traditional evaluation methods [3][8]. - The evaluation framework consists of three levels: object-level, part-level, and material-subject, allowing for a detailed analysis of 3D generation quality [5][8][9][15][17]. - The first 3D generation leaderboard has been published on HuggingFace, featuring 30 mainstream and cutting-edge models, covering both Text-to-3D and Image-to-3D tasks [6][21]. Group 2: Evaluation Dimensions - Object-level evaluation focuses on five dimensions: geometric plausibility, geometric detail, texture quality, geometric-texture coherence, and prompt-3D consistency [9][10][12][14]. - Part-level evaluation allows for a more granular assessment of specific components, such as chair legs or armrests, identifying defects in local structures [15][16]. - Material-subject evaluation assesses surface visual and physical properties across four dimensions: detail and complexity, color and saturation, consistency and artifacts, and material plausibility [17]. Group 3: Leaderboard and Industry Impact - The leaderboard provides a standardized assessment of models under uniform conditions, ensuring fairness and referenceability in the evaluation results [21][28]. - The introduction of Hi3DEval aims to establish a systematic and transparent quality evaluation standard for 3D content, facilitating better comparisons and advancements in the industry [28].