US EditionXPENG-Peking University Collaborative Research Accepted by AAAI 2026: Introducing a Novel Visual Token Pruning Framework for Autonomous Driving
FastDriveVLA cuts computational load by 7.5 times in autonomous driving models while maintaining planning accuracy, advancing scalable deployment for next-generation vehicle autonomy.
- On Dec. 28, 2025, XPENG and Peking University had their paper accepted by AAAI 2026, which accepted 4,167 of 23,680 submissions for a 17.6% rate.
- Because VLA models encode many visual tokens, teams sought ways to speed real-time inference since processing large numbers increases computational load and slows onboard performance.
- Using adversarial foreground-background reconstruction, FastDriveVLA reduced 3,249 tokens to 812, achieving nearly 7.5x compute savings on the nuScenes benchmark.
- This accolade reinforces XPENG's momentum toward scalable, in-vehicle VLA deployment, highlighting full-stack in-house capabilities and following prior recognitions at CVPR WAD and AI Day.
- The work positions token pruning as a path to reconcile scene richness with onboard compute limits, as FastDriveVLA could accelerate on-vehicle inference and help scale safer L4 autonomous driving.
32 Articles
32 Articles
InsideNoVA.com
XPENG-Peking University Collaborative Research Accepted by AAAI 2026: Introducing a Novel Visual Token Pruning Framework for Autonomous Driving
XPENG-PKU Research Breakthrough: XPENG, in collaboration with Peking University, has developed FastDriveVLA—a novel visual token pruning framework that enables autonomous driving AI to "drive like a human" by focusing only on essential information, achieving a 7.5x reduction in computational load.Top-Tier…
Summary The "Human Perspective": FastDriveVLA L-Intelligence Emerging: More than an algorithm, a behavior Giant Infrastructure for a Precision IA Comparison: The Evolution of the IA at XPENG Conclusion: The Industrialization of Intelligence In the race for autonomous driving, raw power is no longer enough. The real challenge is not just to create an IA...
XPENG–Peking University Collaborative Research Accepted By AAAI 2026: Introducing A Novel Visual Token Pruning Framework For Autonomous Driving
XPENG-PKU Research Breakthrough: XPENG, in collaboration with Peking University, has developed FastDriveVLA-a novel visual token pruning framework that enables autonomous driving AI to "drive like a human" by focusing only on essential information, achieving a 7.5x reduction in computational load. Top-Tier AI Recognition: The research has been accepted by AAAI 2026, one of the ... [continued] The post XPENG–Peking University Collaborative Resear…
Coverage Details
Bias Distribution
- 42% of the sources lean Left
Factuality
To view factuality data please Upgrade to Premium
