arXiv:2606.02105v1 Announce Type: new 
Abstract: End-to-End Autonomous Driving (E2E-AD) systems have largely converged on predicting intermediate trajectory waypoints, delegating final control to hand-crafted controllers with GPS access. Direct control-signal prediction (outputting throttle, steer and brake in an end-to-end fashion) remains underexplored, and critically, the role of action multimodality in such systems is not well understood. We argue that moving beyond deterministic, single-action outputs is not merely a modelling choice, but a key driver of driving performance, representational quality, and training stability. To validate this, we introduce the Action Diffusion Transformer (ADT), an anchor-free diffusion transformer trained with a MSE objective that natively models the multimodal distribution of plausible driving actions. Rather than committing to a single deterministic command, ADT generates K action candidates and selects the most suitable one at inference via Nearest Neighbour Matching (NNM). Beyond strong benchmark numbers, we show that action multimodality yields measurable benefits in learned representations and behavioral consistency, effects that deterministic architectures cannot replicate. ADT surpasses previous state-of-the-art on the challenging closed-loop Bench2Drive benchmark while achieving ten times lower latency, demonstrating that expressive, multimodal action modelling is both practically efficient and conceptually essential for robust end-to-end driving.

تقدم دراسة جديدة نموذج Action Diffusion Transformer (ADT)، وهو نهج مبتكر للقيادة الذاتية من النهاية إلى النهاية (E2E-AD) يبرز أهمية مخرجات العمل متعددة الأوجه مقارنة بالطرق الحتمية التقليدية. يقوم هذا النموذج بتوليد عدة مرشحين للعمل، مما يعزز أداء القيادة واستقرار التدريب.

Un nuevo estudio presenta el Action Diffusion Transformer (ADT), un enfoque innovador para la conducción autónoma de extremo a extremo (E2E-AD) que enfatiza la importancia de las salidas de acción multimodal sobre los métodos deterministas tradicionales. Este modelo genera múltiples candidatos de acción, mejorando el rendimiento de conducción y la estabilidad del entrenamiento.

Une nouvelle étude présente l'Action Diffusion Transformer (ADT), une approche novatrice pour la conduite autonome de bout en bout (E2E-AD) qui souligne l'importance des sorties d'action multimodales par rapport aux méthodes déterministes traditionnelles. Ce modèle génère plusieurs candidats d'action, améliorant ainsi la performance de conduite et la stabilité de l'entraînement.

A new study introduces the Action Diffusion Transformer (ADT), a novel approach to End-to-End Autonomous Driving (E2E-AD) that emphasizes the importance of multimodal action outputs over traditional deterministic methods. This model generates multiple action candidates, enhancing driving performance and training stability.