AIAny
Icon for item

SimWAM: A Simple World Action Model for End-to-End Autonomous Driving

Uses video generation only as a training signal to co-train a pretrained video expert and a lightweight action expert, then discards the video branch at inference to produce a low-latency end-to-end driving planner; enhanced with RL for compositional driving rewards.

Introduction

Why this matters Most world-action models (WAMs) improve driving by imagining future video frames at inference, which adds significant compute and latency. SimWAM flips that trade-off: it uses video generation purely as a training signal so the learned action planner can run independently and efficiently at test time while still retaining benefits of video-based dynamics priors.

Key Findings
  • Training-time video supervision: A pretrained video expert is co-trained with a lightweight action expert under a joint flow-matching objective, shaping the action model with spatio-temporal generative priors without requiring future video synthesis at inference. This means you get the modeling benefits of video prediction without its runtime cost.
  • Isolated attention interface: An attention masking scheme keeps action prediction independent of future-frame tokens, allowing the video branch to be removed after training and leaving a self-contained trajectory planner. So what: inference latency and resource usage drop substantially compared to prior WAMs that generate futures online.
  • Reinforcement learning for compositional reward: Beyond imitation, the action expert is further optimized with RL on a compositional driving objective, improving practical driving metrics rather than only matching trajectories.
  • Empirical performance: Achieves 91.5 PDMS on NAVSIM, outperforms state-of-the-art WAM-based planners with much lower latency, and transfers zero-shot to nuScenes, demonstrating both effectiveness and generalization.
Who it's for and tradeoffs

Great fit if you need a deployable, low-latency end-to-end driving planner that benefits from video-derived motion priors but cannot afford test-time video generation. It is also a good baseline for research that wants modular experiments where the video backbone can be swapped independently of the action expert. Look elsewhere if your application requires explicit future-frame outputs (e.g., visual forecasting for downstream perception) or if you need a tightly coupled joint-video/action model at inference. Also, because the approach relies on a pretrained video expert during training, improvements depend on available video foundation models and their computational cost during training.

Method overview

The architecture uses two separate experts (video and action) that share only a unified attention interface and a joint flow-matching loss. During training the video expert predicts future video latents and the action expert learns to decode actions conditioned on those latents (inverse-dynamics style). At inference the action expert runs standalone, producing trajectories directly; the video expert is removed, saving compute.

Information

  • Websitearxiv.org
  • AuthorsZongchuang Zhao, Xin Zhou, Tianyang Xu, Zhengyang Sun, Kaixuan Zhou, Honglin Li, Dingkang Liang, Xiang Bai
  • Published date2026/08/07

More Items

Turns adapter placement for PEFT on YOLO-family real-time detectors into an auditable constraint-planning problem that emits budgeted target-module plans or calibrated refusals; shows planner-selected RS-LoRA improves mAP and cuts peak training memory in evaluated detectors.

Analyzes how to build effective training environment distributions for multimodal agents and proposes Ability-aware Environment Selection (AES) and Hierarchical Difficulty Curriculum (HDC) to improve diversity and difficulty scheduling, yielding large relative gains in experiments.

Evaluates VLMs' ability to form global spatial awareness from long-horizon egocentric video. Introduces GST-Bench: a VQA benchmark with human-verified questions from 6,790 minutes of synthetic video, reveals a large gap (best zero-shot 42.68 vs human 79.08) and provides GST-Train dataset.