Large vision-language models (VLMs) for autonomous driving (AD) are evolving beyond perception and cognition tasks toward motion planning. However, we identify two critical challenges in this direction: (1) VLMs tend to learn shortcuts by relying heavily on history input information, achieving seemingly strong planning results without genuinely understanding the visual inputs; and (2) the chain-of-thought (COT) reasoning processes are always misaligned with the motion planning outcomes, and how to effectively leverage the complex reasoning capability to enhance planning remains largely underexplored. In this paper, we start from a small-scale domain-specific VLM and propose Drive-R1, designed to bridge the scenario reasoning and motion planning for AD. Drive-R1 first undergoes the supervised finetuning on an elaborate dataset containing both long and short COT data. Drive-R1 is encouraged to reason step-by-step from visual input to final planning decisions. Subsequently, Drive-R1 is trained within a reinforcement learning framework that incentivizes the discovery of reasoning paths that are more informative for planning, guided by rewards based on predicted trajectories and meta actions. Experimental evaluations on the nuScenes and DriveLM-nuScenes benchmarks demonstrate that Drive-R1 achieves superior performance compared to existing state-of-the-art VLMs. We believe that Drive-R1 presents a promising direction for bridging reasoning and planning in AD, offering methodological insights for future research and applications.

Existing benchmarks for Vision-Language Model (VLM) on autonomous driving (AD) primarily assess interpretability through open-form visual question answering (QA) within coarse-grained tasks, which remain insufficient to assess capabilities in complex driving scenarios. To this end, we introduce $\textbf{VLADBench}$, a challenging and fine-grained dataset featuring close-form QAs that progress from static foundational knowledge and elements to advanced reasoning for dynamic on-road situations. The elaborate $\textbf{VLADBench}$ spans 5 key domains: Traffic Knowledge Understanding, General Element Recognition, Traffic Graph Generation, Target Attribute Comprehension, and Ego Decision-Making and Planning. These domains are further broken down into 11 secondary aspects and 29 tertiary tasks for a granular evaluation. A thorough assessment of general and domain-specific (DS) VLMs on this benchmark reveals both their strengths and critical limitations in AD contexts. To further exploit the cognitive and reasoning interactions among the 5 domains for AD understanding, we start from a small-scale VLM and train the DS models on individual domain datasets (collected from 1.4M DS QAs across public sources). The experimental results demonstrate that the proposed benchmark provides a crucial step toward a more comprehensive assessment of VLMs in AD, paving the way for the development of more cognitively sophisticated and reasoning-capable AD systems.

Survival prediction of esophageal cancer is an essential task for doctors to make personalized cancer treatment plans. However, handcrafted features from medical images need prior medical knowledge, which is usually limited and not complete, yielding unsatisfying survival predictions. To address these challenges, we propose a novel and efficient deep learning-based survival prediction framework for evaluating clinical outcomes before concurrent chemoradiotherapy. The proposed model consists of two key components: a 3D Coordinate Attention Convolutional Autoencoder (CACA) and an uncertainty-based jointly Optimizing Cox Model (UOCM). The CACA is built upon an autoencoder structure with 3D coordinate attention layers, capturing latent representations and encoding 3D spatial characteristics with precise positional information. Additionally, we designed an Uncertainty-based jointly Optimizing Cox Model, which jointly optimizes the CACA and survival prediction task. The survival prediction task models the interactions between a patient's feature signatures and clinical outcome to predict a reliable hazard ratio of patients. To verify the effectiveness of our model, we conducted extensive experiments on a dataset including computed tomography of 285 patients with esophageal cancer. Experimental results demonstrated that the proposed method achieved a C-index of 0.72, outperforming the state-of-the-art method.