Zero-shot visual question answering (VQA) requires models to reason over unseen image–question pairs without task-specific supervision, where performance is often undermined by incomplete visual grounding and unstable reasoning. Recent approaches attempt to mitigate this by using large language models to generate external knowledge conditioned on captions and questions. However, these methods typically operate in an open-loop manner, lacking mechanisms to regulate competing interpretations or correct misaligned knowledge. Inspired by the hippocampal pattern completion mechanism, which supports inference from partial observations through memory reactivation and competitive stabilization, we propose ARK-PC for zero-shot VQA, which formulates reasoning as a two-stage completion process. It first expands fragmented multimodal cues into multiple structured candidate hypotheses, then adaptively reinforces coherent candidates while suppressing inconsistent ones through iterative feedback. By coupling knowledge generation with competitive refinement, ARK-PC transforms open-loop inference into a closed-loop stabilization process, enabling robust reasoning under uncertainty without external supervision. Experiments on OK-VQA and A-OKVQA demonstrate consistent state-of-the-art zero-shot performance and strong generalization across diverse backbones, indicating that the improvements stem from the proposed framework rather than model scale.

Motor Imagery Brain-Computer Interface (MI-BCI) is a key technology of Brain-Computer Interfaces (BCIs). In practical applications, high-accuracy cross-subject decoding is difficult to achieve due to multi-source heterogeneity caused by individual differences among subjects. Transfer learning technology has been widely applied to try to solve this problem, but existing transfer learning models have neglected the frequency features of EEG signals and the temporal pattern dependency present in high-level features. This restricts the performance and generalization ability of existing cross-subject decoding methods. This study leverages frequency prior knowledge related to motor imagery for data augmentation and designs feature extractor, combined with adversarial learning to achieve cross-subject feature alignment. Unlike traditional adversarial learning frameworks, this work introduces a temporal causal inference module in the classifier to reprocess high-level features, dynamically model the temporal characteristics of EEG signals to reduce bias, and achieve efficient decoding through synchronous training of the discriminator and classifier. Based on a pure convolutional neural network (CNN) architecture, the framework achieved decoding accuracies of 82.64% and 85.51% on the BCI Competition IV 2a and 2b datasets, respectively, surpassing state-of-the-art Transformer architectures (e.g., GAT, BLSAN). These results provide a feasible solution for developing efficient and practical MI-BCI systems and lay a foundation for optimizing cross-subject decoding.

Accurate medical image segmentation is an essential part of the medical image analysis process that provides detailed quantitative metrics. In recent years, extensions of classical networks such as UNet have achieved state-of-the-art performance on medical image segmentation tasks. However, the high model complexity of these networks limits their applicability to devices with constrained computational resources. To alleviate this problem, we propose a shallow hierarchical Transformer for medical image segmentation, called SHFormer. By decreasing the number of transformer blocks utilized, the model complexity of SHFormer can be reduced to an acceptable level. To improve the learned attention while keeping the structure lightweight, we propose a spatial-channel connection module. This module separately learns attention in the spatial and channel dimensions of the feature while interconnecting them to produce more focused attention. To keep the decoder lightweight, the MLP-D module is proposed to progressively fuse multi-scale features in which channels are aligned using Multi-Layer Perceptron (MLP) and spatial information is fused by convolutional blocks. We first validated the performance of SHFormer on the ISIC-2018 dataset. Compared to the latest network, SHFormer exhibits comparable performance with 15 times fewer parameters, 30 times lower computational complexity and 5 times higher inference efficiency. To test the generalizability of SHFormer, we introduced the polyp dataset for additional testing. SHFormer achieves comparable segmentation accuracy to the latest network while having lower computational overhead.