MERGETUNE: Continued fine-tuning of vision-language models

Fine-tuning vision-language models (VLMs) such as CLIP often leads to catastrophic forgetting of pretrained knowledge. Prior work primarily aims to mitigate forgetting during adaptation; however, forgetting often remains inevitable during this process. We introduce a novel paradigm, continued fine-tuning (CFT), which seeks to recover pretrained knowledge after a zero-shot model has already been adapted. We propose a simple, model-agnostic CFT strategy (named MERGETUNE) guided by linear mode connectivity (LMC), which can be applied post hoc to existing fine-tuned models without requiring architectural changes. Given a fine-tuned model, we continue fine-tuning its trainable parameters (e.g., soft prompts or linear heads) to search for a continued model which has two low-loss paths to the zero-shot (e.g., CLIP) and the fine-tuned (e.g., CoOp) solutions. By exploiting the geometry of the loss landscape, the continued model implicitly merges the two solutions, restoring pretrained knowledge lost in the fine-tuned counterpart. A challenge is that the vanilla LMC constraint requires data replay from the pretraining task. We approximate this constraint for the zero-shot model via a second-order surrogate, eliminating the need for large-scale data replay. Experiments show that MERGETUNE improves the harmonic mean of CoOp by +5.6% on base-novel generalisation without adding parameters. On robust fine-tuning evaluations, the LMC-merged model from MERGETUNE surpasses ensemble baselines with lower inference cost, achieving further gains and state-of-the-art results when ensembled with the zero-shot model. Our code is available at https://github.com/Surrey-UP-Lab/MERGETUNE.

Dynamic Avatar-Scene Rendering from Human-centric Context

Reconstructing dynamic humans interacting with real-world environments from monocular videos is an important and challenging task. Despite considerable progress in 4D neural rendering, existing approaches either model dynamic scenes holistically or model scenes and backgrounds separately aim to introduce parametric human priors. However, these approaches either neglect distinct motion characteristics of various components in scene especially human, leading to incomplete reconstructions, or ignore the information exchange between the separately modeled components, resulting in spatial inconsistencies and visual artifacts at human-scene boundaries. To address this, we propose {\bf Separate-then-Map} (StM) strategy that introduces a dedicated information mapping mechanism to bridge separately defined and optimized models. Our method employs a shared transformation function for each Gaussian attribute to unify separately modeled components, enhancing computational efficiency by avoiding exhaustive pairwise interactions while ensuring spatial and visual coherence between humans and their surroundings. Extensive experiments on monocular video datasets demonstrate that StM significantly outperforms existing state-of-the-art methods in both visual quality and rendering accuracy, particularly at challenging human-scene interaction boundaries.

MAGIC-Talk: Motion-aware Audio-Driven Talking Face Generation with Customizable Identity Control

Audio-driven talking face generation has gained significant attention for applications in digital media and virtual avatars. While recent methods improve audio-lip synchronization, they often struggle with temporal consistency, identity preservation, and customization, especially in long video generation. To address these issues, we propose MAGIC-Talk, a one-shot diffusion-based framework for customizable and temporally stable talking face generation. MAGIC-Talk consists of ReferenceNet, which preserves identity and enables fine-grained facial editing via text prompts, and AnimateNet, which enhances motion coherence using structured motion priors. Unlike previous methods requiring multiple reference images or fine-tuning, MAGIC-Talk maintains identity from a single image while ensuring smooth transitions across frames. Additionally, a progressive latent fusion strategy is introduced to improve long-form video quality by reducing motion inconsistencies and flickering. Extensive experiments demonstrate that MAGIC-Talk outperforms state-of-the-art methods in visual quality, identity preservation, and synchronization accuracy, offering a robust solution for talking face generation.

MARS2 2025 Challenge on Multimodal Reasoning: Datasets, Methods, Results, Discussion, and Outlook

This paper reviews the MARS2 2025 Challenge on Multimodal Reasoning. We aim to bring together different approaches in multimodal machine learning and LLMs via a large benchmark. We hope it better allows researchers to follow the state-of-the-art in this very dynamic area. Meanwhile, a growing number of testbeds have boosted the evolution of general-purpose large language models. Thus, this year's MARS2 focuses on real-world and specialized scenarios to broaden the multimodal reasoning applications of MLLMs. Our organizing team released two tailored datasets Lens and AdsQA as test sets, which support general reasoning in 12 daily scenarios and domain-specific reasoning in advertisement videos, respectively. We evaluated 40+ baselines that include both generalist MLLMs and task-specific models, and opened up three competition tracks, i.e., Visual Grounding in Real-world Scenarios (VG-RS), Visual Question Answering with Spatial Awareness (VQA-SA), and Visual Reasoning in Creative Advertisement Videos (VR-Ads). Finally, 76 teams from the renowned academic and industrial institutions have registered and 40+ valid submissions (out of 1200+) have been included in our ranking lists. Our datasets, code sets (40+ baselines and 15+ participants' methods), and rankings are publicly available on the MARS2 workshop website and our GitHub organization page https://github.com/mars2workshop/, where our updates and announcements of upcoming events will be continuously provided.

GrFormer: A Novel Transformer on Grassmann Manifold for Infrared and Visible Image Fusion

In the field of image fusion, promising progress has been made by modeling data from different modalities as linear subspaces. However, in practice, the source images are often located in a non-Euclidean space, where the Euclidean methods usually cannot encapsulate the intrinsic topological structure. Typically, the inner product performed in the Euclidean space calculates the algebraic similarity rather than the semantic similarity, which results in undesired attention output and a decrease in fusion performance. While the balance of low-level details and high-level semantics should be considered in infrared and visible image fusion task. To address this issue, in this paper, we propose a novel attention mechanism based on Grassmann manifold for infrared and visible image fusion (GrFormer). Specifically, our method constructs a low-rank subspace mapping through projection constraints on the Grassmann manifold, compressing attention features into subspaces of varying rank levels. This forces the features to decouple into high-frequency details (local low-rank) and low-frequency semantics (global low-rank), thereby achieving multi-scale semantic fusion. Additionally, to effectively integrate the significant information, we develop a cross-modal fusion strategy (CMS) based on a covariance mask to maximise the complementary properties between different modalities and to suppress the features with high correlation, which are deemed redundant. The experimental results demonstrate that our network outperforms SOTA methods both qualitatively and quantitatively on multiple image fusion benchmarks. The codes are available at https://github.com/Shaoyun2023.

Hybrid Batch Normalisation: Resolving the Dilemma of Batch Normalisation in Federated Learning

Batch Normalisation (BN) is widely used in conventional deep neural network training to harmonise the input-output distributions for each batch of data. However, federated learning, a distributed learning paradigm, faces the challenge of dealing with non-independent and identically distributed data among the client nodes. Due to the lack of a coherent methodology for updating BN statistical parameters, standard BN degrades the federated learning performance. To this end, it is urgent to explore an alternative normalisation solution for federated learning. In this work, we resolve the dilemma of the BN layer in federated learning by developing a customised normalisation approach, Hybrid Batch Normalisation (HBN). HBN separates the update of statistical parameters (i.e. , means and variances used for evaluation) from that of learnable parameters (i.e. , parameters that require gradient updates), obtaining unbiased estimates of global statistical parameters in distributed scenarios. In contrast with the existing solutions, we emphasise the supportive power of global statistics for federated learning. The HBN layer introduces a learnable hybrid distribution factor, allowing each computing node to adaptively mix the statistical parameters of the current batch with the global statistics. Our HBN can serve as a powerful plugin to advance federated learning performance. It reflects promising merits across a wide range of federated learning settings, especially for small batch sizes and heterogeneous data.

One Model for ALL: Low-Level Task Interaction Is a Key to Task-Agnostic Image Fusion

Advanced image fusion methods mostly prioritise high-level missions, where task interaction struggles with semantic gaps, requiring complex bridging mechanisms. In contrast, we propose to leverage low-level vision tasks from digital photography fusion, allowing for effective feature interaction through pixel-level supervision. This new paradigm provides strong guidance for unsupervised multimodal fusion without relying on abstract semantics, enhancing task-shared feature learning for broader applicability. Owning to the hybrid image features and enhanced universal representations, the proposed GIFNet supports diverse fusion tasks, achieving high performance across both seen and unseen scenarios with a single model. Uniquely, experimental results reveal that our framework also supports single-modality enhancement, offering superior flexibility for practical applications. Our code will be available at https://github.com/AWCXV/GIFNet.

UASTrack: A Unified Adaptive Selection Framework with Modality-Customization in Single Object Tracking

Multi-modal tracking is essential in single-object tracking (SOT), as different sensor types contribute unique capabilities to overcome challenges caused by variations in object appearance. However, existing unified RGB-X trackers (X represents depth, event, or thermal modality) either rely on the task-specific training strategy for individual RGB-X image pairs or fail to address the critical importance of modality-adaptive perception in real-world applications. In this work, we propose UASTrack, a unified adaptive selection framework that facilitates both model and parameter unification, as well as adaptive modality discrimination across various multi-modal tracking tasks. To achieve modality-adaptive perception in joint RGB-X pairs, we design a Discriminative Auto-Selector (DAS) capable of identifying modality labels, thereby distinguishing the data distributions of auxiliary modalities. Furthermore, we propose a Task-Customized Optimization Adapter (TCOA) tailored to various modalities in the latent space. This strategy effectively filters noise redundancy and mitigates background interference based on the specific characteristics of each modality. Extensive comparisons conducted on five benchmarks including LasHeR, GTOT, RGBT234, VisEvent, and DepthTrack, covering RGB-T, RGB-E, and RGB-D tracking scenarios, demonstrate our innovative approach achieves comparative performance by introducing only additional training parameters of 1.87M and flops of 1.95G. The code will be available at https://github.com/wanghe/UASTrack.

PortraitTalk: Towards Customizable One-Shot Audio-to-Talking Face Generation

Audio-driven talking face generation is a challenging task in digital communication. Despite significant progress in the area, most existing methods concentrate on audio-lip synchronization, often overlooking aspects such as visual quality, customization, and generalization that are crucial to producing realistic talking faces. To address these limitations, we introduce a novel, customizable one-shot audio-driven talking face generation framework, named PortraitTalk. Our proposed method utilizes a latent diffusion framework consisting of two main components: IdentityNet and AnimateNet. IdentityNet is designed to preserve identity features consistently across the generated video frames, while AnimateNet aims to enhance temporal coherence and motion consistency. This framework also integrates an audio input with the reference images, thereby reducing the reliance on reference-style videos prevalent in existing approaches. A key innovation of PortraitTalk is the incorporation of text prompts through decoupled cross-attention mechanisms, which significantly expands creative control over the generated videos. Through extensive experiments, including a newly developed evaluation metric, our model demonstrates superior performance over the state-of-the-art methods, setting a new standard for the generation of customizable realistic talking faces suitable for real-world applications.

Adaptive Hyper-Graph Convolution Network for Skeleton-based Human Action Recognition with Virtual Connections

The shared topology of human skeletons motivated the recent investigation of graph convolutional network (GCN) solutions for action recognition. However, the existing GCNs rely on the binary connection of two neighbouring vertices (joints) formed by an edge (bone), overlooking the potential of constructing multi-vertex convolution structures. In this paper we address this oversight and explore the merits of a hyper-graph convolutional network (Hyper-GCN) to achieve the aggregation of rich semantic information conveyed by skeleton vertices. In particular, our Hyper-GCN adaptively optimises multi-scale hyper-graphs during training, revealing the action-driven multi-vertex relations. Besides, virtual connections are often designed to support efficient feature aggregation, implicitly extending the spectrum of dependencies within the skeleton. By injecting virtual connections into hyper-graphs, the semantic clues of diverse action categories can be highlighted. The results of experiments conducted on the NTU-60, NTU-120, and NW-UCLA datasets, demonstrate the merits of our Hyper-GCN, compared to the state-of-the-art methods. Specifically, we outperform the existing solutions on NTU-120, achieving 90.2\% and 91.4\% in terms of the top-1 recognition accuracy on X-Sub and X-Set.