Leveraging social homophily to enhance user preference modeling, social recommendation has become a cornerstone of modern recommender systems. However, the raw social network contains inherent unreliability as it teems with noise---misclicks, bot-generated and transient ties---while many meaningful links remain unobserved. In this study, we propose DRSoRec, a dual-rectification model to rectify the raw social networks by simultaneously removing noisy signals and preserving useful information. Specifically, the invariant social rationale discovery module distills each user's influential core social circle of the current recommendation, whereas the adaptive social connection refinement module employs a mixture-of-experts structure learner to prune spurious edges and uncover latent links. A contrastive optimization objective is designed to align and mutually enhance these two modules, and the refined user representations are fused with collaborative representations generated from interactions for the final recommendation. Experiments on three public datasets confirm that DRSoRec consistently gains over state-of-the-art baselines.

Multimodal sentiment analysis (MSA) is attracting increasing attention from researchers. Existing studies on MSA typically rely on surface-level feature extraction and fusion that can be directly obtained from multimodal data, which may often ignore the underlying semantic connection between images and texts. Recent progress in large multimodal models (LMMs) has demonstrated their impressive reasoning abilities, which can be leveraged to improve traditional MSA approaches by providing a deeper understanding of the sematic connection of the modalities. Toward this issue, in this paper, we propose a novel framework called MPT that combines traditional MSA approaches with Multi-Perspective Thinking from LMMs to improve prediction outcomes. Specifically, MPT instructs the traditional multimodal deep learning models to understand multiple-perspective rationales for different sentiment polarities, augmenting its knowledge base and enhancing its ability to make more accurate predictions. Extensive experiments on four refined datasets show that MPT can not only deliver better performance compared with existing methods, but also demonstrate good cross-modal understanding ability for recognizing user sentiment. The codes and datasets can be accessed here: https://github.com/RMJHQwQ/MPT.

Compositional Zero-Shot Learning (CZSL) aims to recognize unseen state-object combinations by leveraging known combinations. Existing studies basically rely on the cross-modal alignment capabilities of CLIP but tend to overlook its limitations in capturing fine-grained local features, which arise from its architectural and training paradigm. To address this issue, we propose a Multi-Stage Cross-modal Interaction (MSCI) model that effectively explores and utilizes intermediate-layer information from CLIP's visual encoder. Specifically, we design two self-adaptive aggregators to extract local information from low-level visual features and integrate global information from high-level visual features, respectively. These key information are progressively incorporated into textual representations through a stage-by-stage interaction mechanism, significantly enhancing the model’s perception capability for fine-grained local visual information. Additionally, MSCI dynamically adjusts the attention weights between global and local visual information based on different combinations, as well as different elements within the same combination, allowing it to flexibly adapt to diverse scenarios. Experiments on three widely used datasets fully validate the effectiveness and superiority of the proposed model. Data and code are available at https://github.com/ltpwy/MSCI.