VLA models have emerged as a powerful paradigm for transferring semantic knowledge from web-scale data to physical robotic control.However, current single-frame architectures suffer from intrinsic limitations: temporal myopia that discards historical dynamics, reasoning gaps between high-level instructions and low-level motor commands, and inference inefficiency due to autoregressive scalar decoding.In this work, we propose MIRTH, a unified framework designed to address these challenges.MIRTH augments a pretrained VLA backbone with three key innovations: (1) dualscale temporal memory hubs that compress long-term scene evolution and short-term motion trends into compact embeddings; (2) latent reasoning tokens optimized via a mutualinformation objective carving out a semantic plan space to align multimodal context with action trajectories; and (3) a parallel action decoding scheme that replaces autoregressive generation with vector-wise prediction to maximize control throughput.Extensive evaluations on the LIBERO simulation benchmark and a real-world LeRobot platform demonstrate that MIRTH achieves state-of-the-art performance and exhibiting emergent error recovery capabilities.The codes and collected datasets are released at http://github.com/kiva12138/mirth.

Semi-supervised learning (SSL) for medical image segmentation is a challenging yet highly practical task, which reduces reliance on large-scale labeled datasets by leveraging unlabeled samples. Among SSL techniques, the weak-to-strong consistency framework, popularized by FixMatch, has emerged as a state-of-the-art method in classification tasks. Notably, such a simple pipeline has also shown competitive performance in medical image segmentation. However, two key limitations still persist, impeding its efficient adaptation: (1) the neglect of contextual dependencies results in inconsistent predictions for similar semantic features, leading to incomplete object segmentation; (2) the lack of exploitation on semantic similarity between labeled and unlabeled data induces considerable class-distribution discrepancy. To address these limitations, we propose a novel SSL framework for medical image segmentation, named S2Match, powered by two appealing designs from a semantic similarity perspective: (1) rectifying pixel-wise prediction by reasoning about the intra-image pair-wise affinity map, thus integrating contextual dependencies explicitly into the final prediction; (2) bridging labeled and unlabeled data via a feature querying mechanism for compact class representation learning, which fully considers cross-image anatomical similarities. As the reliable semantic similarity extraction depends on robust features, we further introduce an effective Spatial-aware Fusion Module (SFM) to explore distinctive information from multiple scales. Experiments show that S2Match yields consistent improvements over the state-of-the-art methods across five public medical image segmentation benchmarks, exhibiting competitive performance on both 2D and 3D tasks.

Deep learning models for medical image segmentation often struggle with task-specific characteristics, limiting their generalization to unseen tasks with new anatomies, labels, or modalities. Retraining or fine-tuning these models requires substantial human effort and computational resources. To address this, in-context learning (ICL) has emerged as a promising paradigm, enabling query image segmentation by conditioning on example image-mask pairs provided as prompts. Unlike previous approaches that rely on implicit modeling or non-end-to-end pipelines, we redefine the core interaction mechanism in ICL as an explicit retrieval process, termed E-ICL, benefiting from the emergence of vision foundation models (VFMs). E-ICL captures dense correspondences between queries and prompts at minimal learning cost and leverages them to dynamically weight multi-class prompt masks. Built upon E-ICL, we propose EICSeg, the first end-to-end ICL framework that integrates complementary VFMs for universal medical image segmentation. Specifically, we introduce a lightweight SD-Adapter to bridge the distinct functionalities of the VFMs, enabling more accurate segmentation predictions. To fully exploit the potential of EICSeg, we further design a scalable self-prompt training strategy and an adaptive token-to-image prompt selection mechanism, facilitating both efficient training and inference. EICSeg is trained on 47 datasets covering diverse modalities and segmentation targets. Experiments on nine unseen datasets demonstrate its strong few-shot generalization ability, achieving an average Dice score of 74.0%, outperforming existing in-context and few-shot methods by 4.5%, and reducing the gap to task-specific models to 10.8%. Even with a single prompt, EICSeg achieves a competitive average Dice score of 60.1%. Notably, it performs automatic segmentation without manual prompt engineering, delivering results comparable to interactive models while requiring minimal labeled data. Source code will be available at https://github.com/zerone-fg/EICSeg.

Understanding human perceptions poses a significant multimodal challenge for computers, involving textual, acoustic, and visual signals. Recently, large language models (LLMs) have garnered great attention, leading to numerous methods aimed at efficiently fine-tuning pretrained models for multimodal downstream tasks. However, there remains a scarcity of techniques that prioritize modality-invariant and -specific information during parameter-efficient tuning, despite evidence from previous studies showcasing the effectiveness of modality disentangling. To address this gap, we propose a novel multimodal tuning approach for LLMs, termed Disentangled Multimodal Tuning and Interaction. Specifically, we evaluate the independence among different modalities and disentangle corresponding modality-invariant and specific components, which are subsequently leveraged for prompt tuning. Following tuning, a newly designed independence-guided cross-attention module is introduced for modality interaction, where the attention mechanism is decoupled and bolstered with independence from the modality-disentangling process. This approach not only enables LLMs to efficiently assimilate information from various modalities but also cultivates an awareness of both modality-invariant and specific information. Compared to previous methods, our approach facilitates modality interaction at a more granular level, resulting in enhanced performance. We validate our method through experiments on four public datasets, demonstrating significant performance improvements.

Medical image segmentation is a crucial yet challenging task in image analysis across diverse anatomical structures. Current segmentation models heavily depend on large-scale datasets, which are laborious to collect and annotate. While generative models offer a promising alternative for data augmentation, most existing approaches are limited to single-modality outputs, either synthetic images or segmentation masks. Moreover, these methods often lack flexible conditioning mechanisms and struggle to capture the rich contextual dependencies inherent in anatomical structures. To address these challenges, in this paper, we propose TPCDM, a novel framework that co-synthesizes high-fidelity paired medical images and segmentation masks through a unified Triple-Prompt Conditional Diffusion Model. At the heart of TPCDM lies a newly defined joint image-label generation paradigm, termed Coordinated Distribution Learning, governed by three synergistic prompts: (1) a text prompt encoding global anatomical semantics; (2) a spatial prompt enforcing pixel-wise spatial coherence; (3) a task prompt dynamically adapting to diverse distributions. Furthermore, TPCDM disentangles instance-wise annotations into semantic masks and distance maps, enabling seamless extension to instance segmentation tasks. Extensive experiments on four benchmarks demonstrate that TPCDM achieves superior synthesis quality. Besides, incorporating the synthesized samples leads to state-of-the-art performance in both downstream semantic and instance segmentation tasks, while also delivering significant improvements under limited labeled data.

In recent years, large-scale pre-trained multimodal models (LMMs) generally emerge to integrate the vision and language modalities, achieving considerable success in multimodal tasks, such as text-image classification. The growing size of LMMs, however, results in a significant computational cost for fine-tuning these models for downstream tasks. Hence, prompt-based interaction strategy is studied to align modalities more efficiently. In this context, we propose a novel efficient prompt-based multimodal interaction strategy, namely Efficient Prompt Interaction for text-image Classification (EPIC). Specifically, we utilize temporal prompts on intermediate layers, and integrate different modalities with similarity-based prompt interaction, to leverage sufficient information exchange between modalities. Utilizing this approach, our method achieves reduced computational resource consumption and fewer trainable parameters (about 1% of the foundation model) compared to other fine-tuning strategies. Furthermore, it demonstrates superior performance on the UPMC-Food101 and SNLI-VE datasets, while achieving comparable performance on the MM-IMDB dataset.

Multimodal sentiment analysis seeks to utilize various types of signals to identify underlying emotions and sentiments. A key challenge in this field lies in multimodal representation learning, which aims to develop effective methods for integrating multimodal features into cohesive representations. Recent advancements include two notable approaches: one focuses on decomposing multimodal features into modality-invariant and -specific components, while the other emphasizes the use of mutual information to enhance the fusion of modalities. Both strategies have demonstrated effectiveness and yielded remarkable results. In this paper, we propose a novel learning framework that combines the strengths of these two approaches, termed mutual information-based disentangled multimodal representation learning. Our approach involves estimating different types of information during feature extraction and fusion stages. Specifically, we quantitatively assess and adjust the proportions of modality-invariant, -specific, and -complementary information during feature extraction. Subsequently, during fusion, we evaluate the amount of information retained by each modality in the fused representation. We employ mutual information or conditional mutual information to estimate each type of information content. By reconciling the proportions of these different types of information, our approach achieves state-of-the-art performance on popular sentiment analysis benchmarks, including CMU-MOSI and CMU-MOSEI.

Single-domain generalization (SDG) can efficiently enhance model generalization while avoiding high annotation costs and privacy concerns. However, existing SDG methods are mainly based on data manipulation and meta-learning, which are not efficient enough due to the limited generalization performance and complex inference. In response to these challenges, we present a novel single domaininvariant representation learning approach for medical image segmentation, called IRLSG, with two appealing designs: (1) A Classscale Photo-metric Augmentation is first proposed to simulate unseen target domain that is sufficient in diversity and informativeness. After that, a Dual-Consistency Framework is further designed to constrain the consistency of intermediate features and segmentation results between the original and the augmented images, which helps to explore the domain-invariant representation. (2) A simple and effective Style Feature Whitening is designed to decouple and remove the domain-specific style from higher-order covariance statistics, which can further improve the modeling and generalization capability of the network. Experimental results on different benchmarks demonstrate that our IRLSG outperforms the current state-of-the-art methods in tackling single-domain generalization.

Domain generalization (DG) aims to generalize the knowledge learned from multiple source domains to unseen target domains. Existing DG techniques can be subsumed under two broad categories, i.e., domain-invariant representation learning and domain manipulation. Nevertheless, it is extremely difficult to explicitly augment or generate the unseen target data. And when source domain variety increases, developing a domain-invariant model by simply aligning more domain-specific information becomes more challenging. In this paper, we propose a simple yet effective method for domain generalization, named Knowledge Distillation based Domain-invariant Representation Learning (KDDRL), that learns domain-invariant representation while encouraging the model to maintain domain-specific features, which recently turned out to be effective for domain generalization. To this end, our method incorporates multiple auxiliary student models and one student leader model to perform a two-stage distillation. In the first-stage distillation, each domain-specific auxiliary student treats the ensemble of other auxiliary students' predictions as a target, which helps to excavate the domain-invariant representation. Also, we present an error removal module to prevent the transfer of faulty information by eliminating incorrect predictions compared to the true labels. In the second-stage distillation, the student leader model with domain-specific features combines the domain-invariant representation learned from the group of auxiliary students to make the final prediction. Extensive experiments and in-depth analysis on popular DG benchmark datasets demonstrate that our KDDRL significantly outperforms the current state-of-the-art methods.

Conventional unsupervised domain adaptation (UDA) and domain generalization (DG) methods rely on the assumption that all source domains can be directly accessed and combined for model training. However, this centralized training strategy may violate privacy policies in many real-world applications. A paradigm for tackling this problem is to train multiple local models and aggregate a generalized central model without data sharing. Recent methods have made remarkable advancements in this paradigm by exploiting parameter alignment and aggregation. But when sources domain variety increases, directly aligning and aggregating local parameters becomes more challenging. Adapting a different approach in this work, we devised a data-free semantic collaborative distillation strategy to learn domain-invariant representation for both federated UDA and DG. Each local model transmits its predictions to the central server and derives its target distribution from the average of other local models' distributions to facilitate the mutual transfer of domain-specific knowledge. When unlabeled target data is available, we introduce a novel UDA strategy termed knowledge filter to adapt the central model to the target data. Extensive experiments on four UDA and DG datasets demonstrate that our method has a competitive performance compared with the state-of-the-art methods.

Semi-supervised learning(SSL) is a data-efficient way in leveraging large-scale data without annotations and alleviating the dependence on labeled data. Mean-Teacher (MT) scheme with teacher-student model architecture has shown its effectiveness in semi-supervised medical image segmentation, where the student network learns from the teacher by minimizing pixel-wise consistency loss. However, existing MT-based SSLs still give rise to two main concerns: (1) limited learning ability of student network that neglects the union of local feature and global cues extraction which may impact the representation learning of variable objects. (2) limited knowledge-transferring ability of teacher network with only pixel-level consistency regularization that may result in inadequate and unstable guidance. To address these limitations, we propose a novel semi-supervised learning scheme, namely MedFCT, with two appealing designs: (1) A dual student architecture with parallel CNN and Transformer branches is designed for local-global feature extraction, where the full-frequency interaction between CNN and Transformer can be explored by a frequency domain cross-fusion (FDCF) module to learn complementarity of the two-paradigm features. (2) A comprehensive multi-level consistency regularization considering pixel-wise, feature-wise and class-wise information is presented to realize more effective guidance and knowledge transfer from teacher network. Experiments show that MedFCT outperforms previous state-of-the-art methods on two public medical image segmentation benchmarks.

Referring image segmentation aims to segment an object out of an image via a specific language expression. The main concept is establishing global visual-linguistic relationships to locate the object and identify boundaries using details of the image. Recently, various Transformer-based techniques have been proposed to efficiently leverage long-range cross-modal dependencies, enhancing performance for referring segmentation. However, existing methods consider visual feature extraction and cross-modal fusion separately, resulting in insufficient visual-linguistic alignment in semantic space. In addition, they employ sequential structures and hence lack multi-scale information interaction. To address these limitations, we propose a Scale-Wise Language-Guided Vision Transformer (SLViT) with two appealing designs: (1) Language-Guided Multi-Scale Fusion Attention, a novel attention mechanism module for extracting rich local visual information and modeling global visual-linguistic relationships in an integrated manner. (2) An Uncertain Region Cross-Scale Enhancement module that can identify regions of high uncertainty using linguistic features and refine them via aggregated multi-scale features. We have evaluated our method on three benchmark datasets. The experimental results demonstrate that SLViT surpasses state-of-the-art methods with lower computational cost. The code is publicly available at: https://github.com/NaturalKnight/SLViT.

The task of multi-label image classification involves recognizing multiple objects within a single image. Considering both valuable semantic information contained in the labels and essential visual features presented in the image, tight visual-linguistic interactions play a vital role in improving classification performance. Moreover, given the potential variance in object size and appearance within a single image, attention to features of different scales can help to discover possible objects in the image. Recently, Transformer-based methods have achieved great success in multi-label image classification by leveraging the advantage of modeling long-range dependencies, but they have several limitations. Firstly, existing methods treat visual feature extraction and cross-modal fusion as separate steps, resulting in insufficient visual-linguistic alignment in the joint semantic space. Additionally, they only extract visual features and perform cross-modal fusion at a single scale, neglecting objects with different characteristics. To address these issues, we propose a Hierarchical Scale-Aware Vision-Language Transformer (HSVLT) with two appealing designs: (1)A hierarchical multi-scale architecture that involves a Cross-Scale Aggregation module, which leverages joint multi-modal features extracted from multiple scales to recognize objects of varying sizes and appearances in images. (2)Interactive Visual-Linguistic Attention, a novel attention mechanism module that tightly integrates cross-modal interaction, enabling the joint updating of visual, linguistic and multi-modal features. We have evaluated our method on three benchmark datasets. The experimental results demonstrate that HSVLT surpasses state-of-the-art methods with lower computational cost.

Domain generalization (DG) demonstrates superior generalization ability in cross-center medical image segmentation. Despite its great success, existing fully supervised DG methods require collecting a large quantity of pixel-level annotations which is quite expensive and time-consuming. To address this challenge, several semi-supervised domain generalized (SSDG) methods have been proposed by simply coupling semi-supervised learning (SSL) with DG tasks, which give rise to two main concerns: (1) Intra-domain dubious semantic information: the quality of pseudo labels in each source domain suffers from the limited amount of labeled data and cross-domain discrepancy. (2) Inter-domain intangible style relationship: current models fail in integrating domain-level information and overlook the relationships among different domains, which degrades the generalization ability of model. In light of these two issues, we propose a novel SSDG framework, namely IS2Net, by arranging an inter-domain generalization branch and several intra-domain SSL branches in a parallel manner, powered by two appealing designs that build a positive interaction between them: (1) A style and semantic memory mechanism is designed to provide both high-quality class-wise representations for intra-domain semantic enhancement and stable domain-specific knowledge for inter-domain style relationship construction. (2) Confident pseudo labeling strategy aims at generating more reliable supervision for intra and inter domain branches, and thus facilitating the learning process of the whole framework. Extensive experiments show that IS2Net yields consistent improvements over the state-of- the-art methods in three public benchmarks.