# FastDriveVLA: Efficient End-to-End Driving via Plug-and-Play Reconstruction-based Token Pruning

arXiv: https://arxiv.org/abs/2507.23318 · HTML: https://mlsanity.com/paper/2507.23318/

## Key facts

- Cited by 20 papers in the map (data through 14 Oct 2026).
- Placed in VLA for autonomous driving, one of 13 directions in Vision-language-action models.
- Builds on 33 of its 46 references that are in the map.
- First posted on arXiv 31 Jul 2025.

## Placed in

- Vision-language-action models / [VLA for autonomous driving](https://mlsanity.com/vla/vla-for-autonomous-driving/)
- End-to-end autonomous driving / [VLMs for autonomous driving](https://mlsanity.com/driving/vlms-for-autonomous-driving/)
- Efficient LLM inference and serving / [Visual token pruning in multimodal models](https://mlsanity.com/inference/visual-token-pruning-in-multimodal-models/)

## Builds on (most cited first, 30 of 33)

- [Masked Autoencoders Are Scalable Vision Learners](https://mlsanity.com/paper/2111.06377/) (2021)
- [Generative Adversarial Networks](https://mlsanity.com/paper/2203.00667/) (2022)
- [Qwen2.5-VL Technical Report](https://mlsanity.com/paper/2502.13923/) (2025)
- [nuScenes: A multimodal dataset for autonomous driving](https://mlsanity.com/paper/1903.11027/) (2019)
- [Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution](https://mlsanity.com/paper/2409.12191/) (2024)
- [Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation](https://mlsanity.com/paper/1308.3432/) (2013)
- [PointPillars: Fast Encoders for Object Detection from Point Clouds](https://mlsanity.com/paper/1812.05784/) (2018)
- [SimMIM: A Simple Framework for Masked Image Modeling](https://mlsanity.com/paper/2111.09886/) (2021)
- [End to End Learning for Self-Driving Cars](https://mlsanity.com/paper/1604.07316/) (2016)
- [Vision Transformers Need Registers](https://mlsanity.com/paper/2309.16588/) (2023)
- [VAD: Vectorized Scene Representation for Efficient Autonomous Driving](https://mlsanity.com/paper/2303.12077/) (2023)
- [Large Scale Interactive Motion Forecasting for Autonomous Driving : The Waymo Open Motion Dataset](https://mlsanity.com/paper/2104.10133/) (2021)
- [NuPlan: A closed-loop ML-based planning benchmark for autonomous vehicles](https://mlsanity.com/paper/2106.11810/) (2021)
- [SparseVLM: Visual Token Sparsification for Efficient Vision-Language Model Inference](https://mlsanity.com/paper/2410.04417/) (2024)
- [Multi-Modal Fusion Transformer for End-to-End Autonomous Driving](https://mlsanity.com/paper/2104.09224/) (2021)
- [DenseTNT: End-to-end Trajectory Prediction from Dense Goal Sets](https://mlsanity.com/paper/2108.09640/) (2021)
- [PyramidDrop: Accelerating Your Large Vision-Language Models via Pyramid Visual Redundancy Reduction](https://mlsanity.com/paper/2410.17247/) (2024)
- [Trajectory-guided Control Prediction for End-to-end Autonomous Driving: A Simple yet Strong Baseline](https://mlsanity.com/paper/2206.08129/) (2022)
- [Multimodal Motion Prediction with Stacked Transformers](https://mlsanity.com/paper/2103.11624/) (2021)
- [Honeybee: Locality-enhanced Projector for Multimodal LLM](https://mlsanity.com/paper/2312.06742/) (2023)
- [End-to-End Urban Driving by Imitating a Reinforcement Learning Coach](https://mlsanity.com/paper/2108.08265/) (2021)
- [OpenDriveVLA: Towards End-to-end Autonomous Driving with Large Vision Language Action Model](https://mlsanity.com/paper/2503.23463/) (2025)
- [TokenPacker: Efficient Visual Projector for Multimodal LLM](https://mlsanity.com/paper/2407.02392/) (2024)
- [LLaVA-Mini: Efficient Image and Video Large Multimodal Models with One Vision Token](https://mlsanity.com/paper/2501.03895/) (2025)
- [Beyond Text-Visual Attention: Exploiting Visual Cues for Effective Token Pruning in VLMs](https://mlsanity.com/paper/2412.01818/) (2024)
- [Stop Looking for Important Tokens in Multimodal Language Models: Duplication Matters More](https://mlsanity.com/paper/2502.11494/) (2025)
- [OmniDrive: A Holistic Vision-Language Dataset for Autonomous Driving with Counterfactual Reasoning](https://mlsanity.com/paper/2504.04348/) (2025)
- [Beyond Attention or Similarity: Maximizing Conditional Diversity for Token Pruning in MLLMs](https://mlsanity.com/paper/2506.10967/) (2025)
- [Impromptu VLA: Open Weights and Open Data for Driving Vision-Language-Action Models](https://mlsanity.com/paper/2505.23757/) (2025)
- [PACT: Pruning and Clustering-Based Token Reduction for Faster Visual Language Models](https://mlsanity.com/paper/2504.08966/) (2025)

## Built on by (most cited first, 20 of 20)

- [A Survey on Efficient Vision-Language-Action Models](https://mlsanity.com/paper/2510.24795/) (2025)
- [Vision-Language-Action Models for Autonomous Driving: Past, Present, and Future](https://mlsanity.com/paper/2512.16760/) (2025)
- [Efficient Vision-Language-Action Models for Embodied Manipulation: A Systematic Survey](https://mlsanity.com/paper/2510.17111/) (2025)
- [StreamKV: Streaming Video Question-Answering with Segment-based KV Cache Retrieval and Compression](https://mlsanity.com/paper/2511.07278/) (2025)
- [KERV: Kinematic-Rectified Speculative Decoding for Embodied VLA Models](https://mlsanity.com/paper/2603.01581/) (2026)
- [ParkGaussian: Surround-view 3D Gaussian Splatting for Autonomous Parking](https://mlsanity.com/paper/2601.01386/) (2026)
- [VDRive: Leveraging Reinforced VLA and Diffusion Policy for End-to-end Autonomous Driving](https://mlsanity.com/paper/2510.15446/) (2025)
- [Distilling Cross-Modal Knowledge via Feature Disentanglement](https://mlsanity.com/paper/2511.19887/) (2025)
- [EvoDriveVLA: Evolving Driving VLA Models via Collaborative Perception-Planning Distillation](https://mlsanity.com/paper/2603.09465/) (2026)
- [RoboECC: Multi-Factor-Aware Edge-Cloud Collaborative Deployment for VLA Models](https://mlsanity.com/paper/2603.20711/) (2026)
- [DRIVESPATIAL: A Benchmark for Spatiotemporal Intelligence in VLMs for Autonomous Driving](https://mlsanity.com/paper/2605.23176/) (2026)
- [TOPS: First-Principles Visual Token Pruning via Constructing Token Optimal Preservation Sets for Efficient MLLM Inference](https://mlsanity.com/paper/2606.27161/) (2026)
- [Efficient Token Pruning for LLaDA-V](https://mlsanity.com/paper/2601.20168/) (2026)
- [ST-Prune: Training-Free Spatio-Temporal Token Pruning for Vision-Language Models in Autonomous Driving](https://mlsanity.com/paper/2604.19145/) (2026)
- [GraspFoM: Towards Reconstruction-Driven Robotic Grasping with 3D Foundation Priors](https://mlsanity.com/paper/2606.08440/) (2026)
- [Efficient Visual Question Answering Pipeline for Autonomous Driving via Scene Region Compression](https://mlsanity.com/paper/2601.07092/) (2026)
- [TPS-Drive: Task-Guided Representation Purification for VLM-based Autonomous Driving](https://mlsanity.com/paper/2605.27038/) (2026)
- [SparseStreet: Sparse Gaussian Splatting for Real-Time Street Scene Simulation](https://mlsanity.com/paper/2606.03909/) (2026)
- [SlackDrive: Reclaiming Runtime Slack for Adaptive Driving Inference](https://mlsanity.com/paper/2609.28064/) (2026)
- [DriveHierarchy: A Benchmark for Diagnosing VLM Driving Capabilities from Open-Loop Understanding to Closed-Loop Execution](https://mlsanity.com/paper/2609.31814/) (2026)

## Abstract (arXiv)

Vision-Language-Action (VLA) models have demonstrated significant potential in complex scene understanding and action reasoning, leading to their increasing adoption in end-to-end autonomous driving systems. However, the long visual tokens of VLA models greatly increase computational costs. Current visual token pruning methods in Vision-Language Models (VLM) rely on either visual token similarity or visual-text attention, but both have shown poor performance in autonomous driving scenarios. Given that human drivers concentrate on relevant foreground areas while driving, we assert that retaining visual tokens containing this foreground information is essential for effective decision-making. Inspired by this, we propose FastDriveVLA, a novel reconstruction-based vision token pruning framework designed specifically for autonomous driving. FastDriveVLA includes a plug-and-play visual token pruner called ReconPruner, which prioritizes foreground information through MAE-style pixel reconstruction. A novel adversarial foreground-background reconstruction strategy is designed to train ReconPruner for the visual encoder of VLA models. Once trained, ReconPruner can be seamlessly applied to different VLA models with the same visual encoder without retraining. To train ReconPruner, we also introduce a large-scale dataset called nuScenes-FG, consisting of 241K image-mask pairs with annotated foreground regions. Our approach achieves state-of-the-art results on the nuScenes open-loop planning benchmark across different pruning ratios.