Computed Tomography (CT) to Cone-Beam Computed Tomography (CBCT) image registration is crucial for image-guided radiotherapy and surgical procedures. However, achieving accurate CT-CBCT registration remains challenging due to various factors such as inconsistent intensities, low contrast resolution and imaging artifacts. In this study, we propose a Context-Aware Semantics-driven Hierarchical Network (referred to as CASHNet), which hierarchically integrates context-aware semantics-encoded features into a coarse-to-fine registration scheme, to explicitly enhance semantic structural perception during progressive alignment. Moreover, it leverages diffeomorphisms to integrate rigid and non-rigid registration within a single end-to-end trainable network, enabling anatomically plausible deformations and preserving topological consistency. CASHNet comprises a Siamese Mamba-based multi-scale feature encoder and a coarse-to-fine registration decoder, which integrates a Rigid Registration (RR) module with multiple Semantics-guided Velocity Estimation and Feature Alignment (SVEFA) modules operating at different resolutions. Each SVEFA module comprises three carefully designed components: i) a cross-resolution feature aggregation (CFA) component that synthesizes enhanced global contextual representations, ii) a semantics perception and encoding (SPE) component that captures and encodes local semantic information, and iii) an incremental velocity estimation and feature alignment (IVEFA) component that leverages contextual and semantic features to update velocity fields and to align features. These modules work synergistically to boost the overall registration performance. Extensive experiments on three typical yet challenging CT-CBCT datasets of both soft and hard tissues demonstrate the superiority of our proposed method over other state-of-the-art methods. The code will be publicly available at https://github.com/xiaorugao999/CASHNet.

Universal medical image registration through a single model handling various registration tasks has attracted increasing interest. However, existing deep learning-based methods face two major challenges in adapting to universal registration tasks: 1) they lack generalizable feature representation capabilities for cross-task registration; 2) they rely solely on model architectures with fixed parameters, which limits their flexibility to dynamically adapt to different registration tasks and inherently compromises their generalization capability for zero-shot performance on unseen tasks. To address these limitations, we propose CIM-VTP, a novel two-stage universal registration framework. In the first stage, our proposed Correlation-guided Image Modeling (CIM)-based pretraining strategy leverages cross-image correlation to guide the masked modeling process, which facilitates spatial correspondence capturing that is essential for registration and provides universal representation capabilities as a foundation for registration learning. In the second stage, we introduce a registration task classifier to identify the type of a given input task, which explicitly quantifies the similarity between current inputs and previously seen tasks. The obtained task similarity scores are then fed as prior information into our carefully designed multi-resolution Visual-Textual Task Prompt (VTP) modules, which integrate task-relevant knowledge through prompt learning to adaptively adjust decoder parameters for different input domains. Extensive experiments across six different registration tasks demonstrate that the proposed CIM-VTP exhibits superior universal image registration performance. The code will be released at https://github.com/xiehousheng/CIM-VTP.

Corporate relative valuation (CRV) refers to the process of comparing a company's value from company products, core staff and other related information, so that we can assess the company's market value, which is critical for venture capital firms. Traditionally, relative valuation methods heavily rely on tedious and expensive human efforts, especially for non-publicly listed companies. However, the availability of information about company's invisible assets, such as patents, talent, and investors, enables a new paradigm for learning and evaluating corporate relative values automatically. Indeed, in this paper, we reveal that, if the companies and their core members are formed as a heterogeneous graph and the attributes of different nodes include semantically-rich multi-modal data, it is able to extract a latent embedding for each company. Along this line, we develop an end-to-end heterogeneous multi-modal graph neural network method, named HM$^2$. Specifically, HM$^2$firstly perform the representation learning for heterogeneous neighbors of input company by taking relationships among nodes into consideration, which aggregates node attributes via linkage-aware multi-head attention mechanism, rather than multi-instance based methods. Then, HM$^2$adopts the self-attention network to aggregate different modal embeddings for final prediction, and employs dynamic triplet loss with embeddings of competitors as the constraint.

In talent management, resume assessment aims to analyze the quality of a job seeker's resume, which can assist recruiters to discover suitable candidates and benefit job seekers improving resume quality in return. Recent machine learning based methods on large-scale public resume datasets have provided the opportunity for automatic assessment for reducing manual costs. However, most existing approaches are still content-dominated and ignore other valuable information. Inspired by practical resume evaluations that consider both the content and layout, we construct the multi-modalities from resumes but face a new challenge that sometimes the performance of multi-modal fusion is even worse than the best uni-modality. In this paper, we experimentally find that this phenomenon is due to the cross-modal divergence. Therefore, we need to consider when is it appropriate to perform multi-modal fusion? To address this problem, we design an instance-aware fusion method, i.e., Divergence-Orientated Multi-Modal Fusion Network (DOMFN), which can adaptively fuse the uni-modal predictions and multi-modal prediction based on cross-modal divergence. Specifically, DOMFN computes a functional penalty score to measure the divergence of cross-modal predictions. Then, the learned divergence can be used to decide whether to conduct multi-modal fusion and be adopted into an amended loss for reliable training. Consequently, DOMFN rejects multi-modal prediction when the cross-modal divergence is too large, avoiding the overall performance degradation, so as to achieve better performance than uni-modalities. In experiments, qualitative comparison with baselines on real-world dataset demonstrates the superiority and explainability of the proposed DOMFN, e.g., we find a meaningful phenomenon that multi-modal fusion has positive effects for assessing resumes from UI Designer and Enterprise Service positions, whereas affects the assessment of Technology and Product Operation positions.