Modern auto-bidding systems are required to balance overall performance with diverse advertiser goals and real-world constraints, reflecting the dynamic and evolving needs of the industry. Recent advances in conditional generative models, such as transformers and diffusers, have enabled direct trajectory generation tailored to advertiser preferences, offering a promising alternative to traditional Markov Decision Process-based methods. However, these generative methods face significant challenges, such as the distribution shift between offline and online environments, limited exploration of the action space, and the necessity to meet constraints like marginal Cost-per-Mille (CPM) and Return on Investment (ROI). To tackle these challenges, we propose GRAD (Generative Reward-driven Ad-bidding with Mixture-of-Experts), a scalable foundation model for auto-bidding that combines an Action-Mixture-of-Experts module for diverse bidding action exploration with the Value Estimator of Causal Transformer for constraint-aware optimization. Extensive offline and online experiments demonstrate that GRAD significantly enhances platform revenue, highlighting its effectiveness in addressing the evolving and diverse requirements of modern advertisers. Furthermore, GRAD has been implemented in multiple marketing scenarios at Meituan, one of the world's largest online food delivery platforms, leading to a 2.18% increase in Gross Merchandise Value (GMV) and 10.68% increase in ROI.

Registration of multimodal images such as optical data and digital elevation models (DEM) is a challenging task due to significant non-linear differences between these images. To address the problem, this paper proposes a new robust approach to integrate optical images and DEMs by using terrain features. In order to detect these features, a simulated image is generated based on the DEM by illuminating it in the geometry of the optical image acquisition, such that typical textures induced by topography are well present as those in the optical image. Hence, the multimodal registration is performed on the optical data and the simulated image to maximize the accuracy of feature matching. The Affine Scale-Invariant Feature Transform (ASIFT) is used to detect keypoints from the two images. Correspondences are matched through the nearest neighbor distance ratio, and outliers are removed using a two-step elimination procedure. The proposed method has been tested on five pairs of optical-DEM images with various spatial resolutions. Experimental results have shown that this method can provide robust registration for optical-DEM images with high accuracies of sub pixel level. The novelty of proposed method is to make use of terrain features for registration, providing a new perspective for the integration of multimodal images.