SimWAM：適用於端到端自動駕駛的簡潔世界動作模型

摘要

世界-動作模型（WAMs）透過將影片動態先驗轉移至動作預測來提升端到端自動駕駛，但現有方法在推論時需要昂貴的未來生成。我們提出 SimWAM，一個簡單而有效的 WAM，將影片生成純粹用作訓練訊號。它利用聯合流匹配共同訓練一個預訓練的影片專家與一個輕量級的動作專家。隔離的注意力遮罩使動作預測獨立於未來幀，從而在訓練後可丟棄影片分支，留下一個可直接預測軌跡的自包含規劃器。由於兩個專家不共享參數，且僅透過統一的注意力介面互動，影片骨幹可被替換，動作專家也可獨立擴展，而無需修改學習目標或推論流程。我們進一步應用強化學習來最佳化超越軌跡模仿的組合式駕駛獎勵。我們的 SimWAM 在 NAVSIM 上達到 91.5 PDMS，以顯著更低的延遲超越基於 WAM 的最先進規劃器，並零樣本遷移至 nuScenes。這些結果使 SimWAM 成為一個簡單而穩健的基線，可輕鬆受益於影片生成技術的進展，實現高效自動駕駛。程式碼與模型權重可在 https://github.com/H-EmbodVis/SimWAM/ 取得。

English

World-Action Models (WAMs) improve end-to-end autonomous driving by transferring video dynamics priors to action prediction, but existing methods require costly future generation at inference. We present SimWAM, a simple yet effective WAM that uses video generation purely as a training signal. It co-trains a pretrained video expert and a lightweight action expert with joint flow matching. An isolated attention mask keeps action prediction independent of future frames, allowing the video branch to be discarded after training and leaving a self-contained planner that directly predicts trajectories. Since the two experts share no parameters and interact only through a unified attention interface, the video backbone could be replaced and the action expert scaled independently without modifying the learning objective or inference pipeline. We further apply reinforcement learning to optimize a compositional driving reward beyond trajectory imitation. Our SimWAM achieves 91.5 PDMS on NAVSIM, surpasses state-of-the-art WAM-based planners with substantially lower latency, and transfers zero-shot to nuScenes. These results position SimWAM as a simple yet solid baseline that could readily benefit from advances in video generation for efficient autonomous driving. The code and model weights are available at https://github.com/H-EmbodVis/SimWAM/