arXiv:2605.31572v1 Announce Type: new 
Abstract: Reasoning is essential for autonomous driving (AD) in long-tail scenarios, where vehicles must apply commonsense knowledge, understand spatial relations, infer agent interactions, and make safe decisions. However, existing AD datasets and benchmarks mainly target perception, prediction, or planning, and provide limited supervision for reasoning over realistic long-tail driving scenes. We introduce nuReasoning, a large-scale real-world dataset and benchmark for reasoning-centric AD. Following the lineage of nuScenes and nuPlan, nuReasoning advances real-world AD datasets and benchmarks toward reasoning in long-tail driving scenarios. The dataset contains 20,000 clips, each 20 seconds long, collected across multiple cities, with synchronized multi-camera images, LiDAR data, HD maps, object annotations, and human-verified reasoning annotations spanning Spatial Reasoning, Decision Reasoning, and Counterfactual Reasoning. Unlike prior datasets that focus primarily on visual question answering, nuReasoning supports both reasoning evaluation and planning evaluation, enabling a direct study of how reasoning supervision affects driving performance. Experiments show that fine-tuning VLMs on nuReasoning substantially improves driving-specific question answering, while incorporating reasoning supervision into VLA training improves planning performance even when textual reasoning outputs are disabled at inference time. These results establish nuReasoning as a foundation for evaluating and improving robust, interpretable, reasoning-driven AD systems in realistic long-tail settings.

تم تقديم nuReasoning كبيانات ومقياس على نطاق واسع يهدف إلى تعزيز قدرات التفكير في سيناريوهات القيادة الذاتية الطويلة. يتضمن هذا المجموع 20,000 مقطع مع صور متعددة الكاميرات المتزامنة، وبيانات LiDAR، وتوضيحات تفكير تم التحقق منها من قبل البشر، مما يعالج القيود المفروضة على مجموعات البيانات الحالية التي تركز بشكل أساسي على الإدراك والتخطيط.

nuReasoning se ha introducido como un conjunto de datos y un benchmark a gran escala destinado a mejorar las capacidades de razonamiento en escenarios de conducción autónoma de cola larga. Este conjunto de datos incluye 20,000 clips con imágenes de múltiples cámaras sincronizadas, datos LiDAR y anotaciones de razonamiento verificadas por humanos, abordando las limitaciones de los conjuntos de datos existentes que se centran principalmente en la percepción y la planificación.

nuReasoning a été introduit comme un ensemble de données et un benchmark à grande échelle visant à améliorer les capacités de raisonnement dans des scénarios de conduite autonome à long terme. Cet ensemble de données comprend 20 000 clips avec des images synchronisées de plusieurs caméras, des données LiDAR et des annotations de raisonnement vérifiées par des humains, répondant aux limites des ensembles de données existants qui se concentrent principalement sur la perception et la planification.

nuReasoning has been introduced as a large-scale dataset and benchmark aimed at enhancing reasoning capabilities in long-tail autonomous driving scenarios. This dataset includes 20,000 clips with synchronized multi-camera images, LiDAR data, and human-verified reasoning annotations, addressing the limitations of existing datasets that primarily focus on perception and planning.