XPeng Motors, in collaboration with Peking University, proposed a new visual token pruning framework FastDriveVLA, and the paper was selected for the international AI conference AAAI 2026

Recently, AAAI 2026 announced the results of paper acceptance. This conference is one of the top international conferences in the field of artificial intelligence. It is reported that AAAI 2026 received a total of 23,680 paper submissions, of which 4,167 papers were accepted, resulting in an acceptance rate of only 17.6%. The paper "FastDriveVLA: Efficient End-to-End Driving via Plug-and-Play Reconstruction-based Token Pruning," jointly completed by XPeng and the National Key Laboratory of Multimedia Information Processing at Peking University, was successfully selected. The main contribution of this paper is the proposal of an efficient visual token pruning framework, FastDriveVLA, specifically tailored for end-to-end autonomous driving VLA models. (Auto has style)