Overview: BEV-LLM is a lightweight 1B model for 3D scene captioning in autonomous driving. It fuses LiDAR and multi-view images using BEVFusion and introduces novel absolute positional encoding for view-specific descriptions. Despite its size, it performs competitively on the nuCaption benchmark, enhancing transparency, safety, and human-AI interaction.
© 2025 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.