Multi-modal large language models (MLLMs) have achieved remarkable capabilities by integrating visual perception with language understanding, enabling applications such as image-grounded dialogue, visual question answering, and scientific analysis. However, most MLLMs adopt a static inference paradigm, encoding the entire image into fixed visual tokens upfront, which limits their ability to iteratively refine understanding or adapt to context during inference. This contrasts sharply with human perception, which is dynamic, selective, and feedback-driven. In this work, we introduce a novel framework for inference-time visual token scaling that enables MLLMs to perform iterative, verifier-guided reasoning over visual content. We formulate the problem as a Markov Decision Process, involving a reasoner that proposes visual actions and a verifier—trained via multi-step Direct Preference Optimization (DPO)—that evaluates these actions and determines when reasoning should terminate. To support this, we present a new dataset, VTS, comprising supervised reasoning trajectories (VTS-SFT) and preference-labeled reasoning comparisons (VTS-DPO). Our method significantly outperforms existing approaches across diverse visual reasoning benchmarks, offering not only improved accuracy but also more interpretable and grounded reasoning processes. These results demonstrate the promise of dynamic inference mechanisms for enabling fine-grained, context-aware visual reasoning in next-generation MLLMs. Code and datasets are publicly released at https://vts-v.github.io/.

Distributed Deep Learning (DDL) is widely used to accelerate deep neural network training for various Web applications. In each iteration of DDL training, each worker synchronizes neural network gradients with other workers. This introduces communication overhead and degrades the scaling performance. In this paper, we propose a recursive model, OSF (Scaling Factor considering Overlap), for estimating the scaling performance of DDL training of neural network models, given the settings of the DDL system. OSF captures two main characteristics of DDL training: the overlap between computation and communication, and the tensor fusion for batching updates. Measurements on a real-world DDL system show that OSF obtains a low estimation error (ranging from 0.5% to 8.4% for different models). Using OSF, we identify the factors that degrade the scaling performance, and propose solutions to effectively mitigate their impacts. Specifically, the proposed adaptive tensor fusion improves the scaling performance by 32.2%∼ 150% compared to the constant tensor fusion buffer size.

Convective storm nowcasting has long been an important issue and has attracted substantial interest. 3D radar images and 3D re-analysis data contain spatiotemporal information of the convective processes. This paper proposes a multi-channel 3D convolutional recurrent neural network (3D-CRN) for convective storm nowcasting, which aims to learn spatiotemporal information directly from these 3D radar and reanalysis data. 3D-CRN is composed of two sub-networks: three multi-channel 3D convolutional networks are used as the front-end spatial sub-networks, and the convLSTM encoder-decoder is constructed as a back-end temporal sub-network. By combing two subnets into one unified network, 3D-CRN can be jointly trained effectively. In order to give forecasts at different lead time simultaneously, we construct a many-to-many encoder-decoder structure to avoid tedious need to train several models respectively. Experimental results show the effectiveness of the proposed method.