Vision-Language-Action (VLA) models have advanced in robotic manipulation, yet practical deployment remains hindered by two key limitations: **1) perceptual redundancy**, where irrelevant visual inputs are processed inefficiently, and **2) superficial instruction-vision alignment**, which hampers semantic grounding of actions. In this paper, we propose **SemanticVLA**, a novel VLA framework that performs Semantic-Aligned Sparsification and Enhancement for Efficient Robotic Manipulation. Specifically: **1)** To sparsify redundant perception while preserving semantic alignment, **Semantic-guided Dual Visual Pruner (SD-Pruner)** performs: Instruction-driven Pruner (ID-Pruner) extracts global action cues and local semantic anchors in SigLIP; Spatial-aggregation Pruner (SA-Pruner) compacts geometry-rich features into task-adaptive tokens in DINOv2. **2)** To exploit sparsified features and integrate semantics with spatial geometry, **Semantic-complementary Hierarchical Fuser (SH-Fuser)** fuses dense patches and sparse tokens across SigLIP and DINOv2 for coherent representation. **3)** To enhance the transformation from perception to action, **Semantic-conditioned Action Coupler (SA-Coupler)** replaces the conventional observation-to-DoF approach, yielding more efficient and interpretable behavior modeling for manipulation tasks. Extensive experiments on simulation and real-world tasks show that SemanticVLA sets a new SOTA in both performance and efficiency. SemanticVLA surpasses OpenVLA on LIBERO benchmark by **21.1%** in success rate, while reducing training cost and inference latency by **3.0×** and **2.7×**.

Unified video and action prediction models hold great potential for robotic manipulation, as future observations offer contextual cues for planning, while actions reveal how interactions shape the environment. However, most existing approaches treat observation and action generation in a monolithic and goal-agnostic manner, often leading to semantically misaligned predictions and incoherent behaviors. To this end, we propose H-GAR, a Hierarchical interaction framework via Goal-driven observation-Action Refinement. To anchor prediction to the task objective, H-GAR first produces a goal observation and a coarse action sketch that outline a high-level route toward the goal. To enable explicit interaction between observation and action under the guidance of the goal observation for more coherent decision-making, we devise two synergistic modules. (1) Goal-Conditioned Observation Synthesizer (GOS) synthesizes intermediate observations based on the coarse-grained actions and the predicted goal observation. (2) Interaction-Aware Action Refiner (IAAR) refines coarse actions into fine-grained, goal-consistent actions by leveraging feedback from the intermediate observations and a Historical Action Memory Bank that encodes prior actions to ensure temporal consistency. By integrating goal grounding with explicit action-observation interaction in a coarse-to-fine manner, H-GAR enables more accurate manipulation. Extensive experiments on both simulation and real-world robotic manipulation tasks demonstrate that H-GAR achieves state-of-the-art performance.

While GUI agents have shown strong performance under explicit and completion instructions, real-world deployment requires aligning with users' more complex implicit intents.In this work, we highlight Hierarchical Implicit Intent Alignment for Personalized GUI Agent (PersonalAlign), a new agent task that requires agents to leverage long-term user records as persistent context to resolve omitted preferences in vague instructions and anticipate latent routines by user state for proactive assistance.To facilitate this study, we introduce AndroidIntent, a benchmark designed to evaluate agents' ability in resolving vague instructions and providing proactive suggestions through reasoning over long-term user records.We annotated 775 user-specific preferences and 215 routines from 20k long-term records across different users for evaluation.Furthermore, we introduce Hierarchical Intent Memory Agent (HIM-Agent), which maintains a continuously updating personal memory and hierarchically organizes user preferences and routines for personalization.Finally, we evaluate a range of GUI agents on AndroidIntent, including GPT-5, Qwen3-VL, and UI-TARS, further results show that HIM-Agent significantly improves both execution and proactive performance by 15.7% and 7.3%.

Emotional understanding and generation are often treated as separate tasks, yet they are inherently complementary and can mutually enhance each other. In this paper, we propose the UniEmo, a unified framework that seamlessly integrates these two tasks. The key challenge lies in the abstract nature of emotions, necessitating the extraction of visual representations beneficial for both tasks. To address this, we propose a hierarchical emotional understanding chain with learnable expert queries that progressively extracts multi-scale emotional features, thereby serving as a foundational step for unification. Simultaneously, we fuse these expert queries and emotional representations to guide the diffusion model in generating emotion-evoking images. To enhance the diversity and fidelity of the generated emotional images, we further introduce the emotional correlation coefficient and emotional condition loss into the fusion process. This step facilitates fusion and alignment for emotional generation guided by the understanding. In turn, we demonstrate that joint training allows the generation component to provide implicit feedback to the understanding part. Furthermore, we propose a novel data filtering algorithm to select high-quality and diverse emotional images generated by the well-trained model, which explicitly feedback into the understanding part. Together, these generation-driven dual feedback processes enhance the model's understanding capacity. Extensive experiments show that UniEmo significantly outperforms state-of-the-art methods in both emotional understanding and generation tasks. The code for the proposed method is available at https://github.com/JiuTian-VL/UniEmo.

Bin Xie, Rui Shao, Gongwei Chen, Kaiwen Zhou, Yinchuan Li, Jie Liu, Min Zhang, Liqiang Nie. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Despite the significant success of imitation learning in robotic manipulation, its application to bimanual tasks remains highly challenging. Existing approaches mainly learn a policy to predict a distant next-best end-effector pose (NBP) and then compute the corresponding joint rotation angles for motion using inverse kinematics. However, they suffer from two important issues: (1) rarely considering the physical robotic structure, which may cause self-collisions or interferences, and (2) overlooking the kinematics constraint, which may result in the predicted poses not conforming to the actual limitations of the robot joints. In this paper, we propose Kinematics enhanced Spatial-TemporAl gRaph Diffuser (KStar Diffuser). Specifically, (1) to incorporate the physical robot structure information into action prediction, KStar Diffuser maintains a dynamic spatial-temporal graph according to the physical bimanual joint motions at continuous timesteps. This dynamic graph serves as the robot-structure condition for denoising the actions; (2) to make the NBP learning objective consistent with kinematics, we introduce the differentiable kinematics to provide the reference for optimizing KStar Diffuser. This module regularizes the policy to predict more reliable and kinematics-aware next end-effector poses. Experimental results show that our method effectively leverages the physical structural information and generates kinematics-aware actions in both simulation and real-world.

First-person video assistants are highly anticipated to enhance our daily lives through online video dialogue. However, existing online video assistants often sacrifice assistant efficacy for real-time efficiency by processing low-frame-rate videos with coarse-grained visual features. To overcome the trade-off between efficacy and efficiency, we propose "Fast & Slow Video-Language Thinker" as onLIne videO assistaNt, LION-FS, achieving real-time, proactive, temporally accurate, and contextually precise responses. LION-FS adopts a two-stage optimization strategy: 1) Fast Path: Routing-Based Response Determination evaluates frame-by-frame whether an immediate response is necessary. To enhance response determination accuracy and handle higher frame-rate inputs efficiently, we employ Token Aggregation Routing to dynamically fuse spatiotemporal features without increasing token numbers, while utilizing Token Dropping Routing to eliminate redundant features, and 2) Slow Path: Multi-granularity Keyframe Augmentation optimizes keyframes during response generation. To provide comprehensive and detailed responses beyond atomic actions constrained by training data, fine-grained spatial features and human-environment interaction features are extracted through multi-granular pooling. They are further integrated into a meticulously designed multimodal Thinking Template to guide more precise response generation. Comprehensive evaluations of online video tasks demonstrate that LION-FS achieves state-of-the-art efficacy and efficiency.

Building an agent that can mimic human behavior patterns to accomplish various open-world tasks is a longterm goal. To enable agents to effectively learn behavioral patterns across diverse tasks, a key challenge lies in modeling the intricate relationships among observations, actions, and language. To this end, we propose Optimus-2, a novel Minecraft agent that incorporates a Multimodal Large Language Model (MLLM) for highlevel planning, alongside a Goal-Observation-Action Conditioned Policy (GOAP) for low-level control. GOAP contains (1) an Action-guided Behavior Encoder that models causal relationships between observations and actions at each timestep, then dynamically interacts with the historical observation-action sequence, consolidating it into fixedlength behavior tokens, and (2) an MLLM that aligns behavior tokens with open-ended language instructions to predict actions auto-regressively. Moreover, we introduce a high-quality Minecraft Goal-Observation-Action (MGOA) dataset, which contains 25,000 videos across 8 atomic tasks, providing about 30M goal-observation-action pairs. The automated construction method, along with the MGOA dataset, can contribute to the community’s efforts to train Minecraft agents. Extensive experimental results demonstrate that Optimus-2 exhibits superior performance across atomic tasks, long-horizon tasks, and open-ended instruction tasks in Minecraft. Please see the project page at https://cybertronagent.github.io/Optimus-2.github.io/.

The research focus of GUI agents is shifting from text-dependent to pure-vision-based approaches, which, though promising, prioritize comprehensive pre-training data collection while neglecting contextual modeling challenges. We probe the characteristics of element and history contextual modeling in GUI agent and summarize: 1) the high-density and loose-relation of element context highlight the existence of many unrelated elements and their negative influence; 2) the high redundancy of history context reveals the inefficient history modeling in current GUI agents. In this work, we propose a context-aware simplification framework for building an efficient and effective GUI Agent, termed SimpAgent. To mitigate potential interference from numerous unrelated elements, we introduce a masking-based element pruning method that circumvents the intractable relation modeling through an efficient masking mechanism. To reduce the redundancy in historical information, we devise a consistency-guided history compression module, which enhances implicit LLM-based compression through innovative explicit guidance, achieving an optimal balance between performance and efficiency. With the above components, SimpAgent reduces 27% FLOPs and achieves superior GUI navigation performances. Comprehensive navigation experiments across diverse web and mobile environments demonstrate the effectiveness and potential of our agent.

The incorporation of high-resolution visual input equips multimodal large language models (MLLMs) with enhanced visual perception capabilities for real-world tasks. However, most existing high-resolution MLLMs rely on a cropping-based approach to process images, which leads to fragmented visual encoding and a sharp increase in redundant tokens. To tackle these issues, we propose the FALCON model. FALCON introduces a novel visual register technique to simultaneously: 1) Eliminate redundant tokens at the stage of visual encoding. To directly address the visual redundancy present in the output of vision encoder, we propose a Register-based Representation Compacting (ReCompact) mechanism. This mechanism introduces a set of learnable visual registers designed to adaptively aggregate essential information while discarding redundancy. It enables the encoder to produce a more compact visual representation with a minimal number of output tokens, thus eliminating the need for an additional compression module. 2) Ensure continuity in visual encoding. To address the potential encoding errors caused by fragmented visual inputs, we develop a Register Interactive Attention (ReAtten) module. This module facilitates effective and efficient information exchange across sub-images by enabling interactions between visual registers. It ensures the continuity of visual semantics throughout the encoding. We conduct comprehensive experiments with FALCON on high-resolution benchmarks across a wide range of scenarios. FALCON demonstrates superior performance with a remarkable 9-fold reduction in visual tokens.

Smartphone agents are increasingly important for helping users control devices efficiently, with (Multimodal) Large Language Model (MLLM)-based approaches emerging as key contenders. Fairly comparing these agents is essential but challenging, requiring a varied task scope, the integration of agents with different implementations, and a generalisable evaluation pipeline to assess their strengths and weaknesses. In this paper, we present SPA-Bench, a comprehensive SmartPhone Agent Benchmark designed to evaluate (M)LLM-based agents in an interactive environment that simulates real-world conditions. SPA-Bench offers three key contributions: (1) A diverse set of tasks covering system and third-party apps in both English and Chinese, focusing on features commonly used in daily routines; (2) A plug-and-play framework enabling real-time agent interaction with Android devices, integrating over ten agents with the flexibility to add more; (3) A novel evaluation pipeline that automatically assesses agent performance across multiple dimensions, encompassing seven metrics related to task completion and resource consumption. Our extensive experiments across tasks and agents reveal challenges like interpreting mobile user interfaces, action grounding, memory retention, and execution costs. We propose future research directions to ease these difficulties, moving closer to real-world smartphone agent applications.

Transforming complex actions into discrete skill abstractions has demonstrated strong potential for robotic manipulation.Existing approaches mainly leverage latent variable models, e.g., VQ-VAE, to learn skill abstractions through learned vectors (codebooks), while they suffer from codebook collapse and modeling the causal relationship between learned skills. To address these limitations, we present **S**kill **T**raining with **A**ugmented **R**otation (**STAR**), a framework that advances both skill learning and composition to complete complex behaviors. Specifically, to prevent codebook collapse, we devise rotation-augmented residual skill quantization (RaRSQ).It encodes relative angles between encoder outputs into the gradient flow by rotation-based gradient mechanism. Points within the same skill code are forced to be either pushed apart or pulled closer together depending on gradient directions.Further, to capture the casual relationship between skills, we present causal skill transformer (CST) which explicitly models dependencies between skill representations through an autoregressive mechanism for coherent action generation.Extensive experiments demonstrate the superiority of STAR on both LIBERO benchmark and realworld tasks, with around 12% improvement over the baselines.

Current affective computing paradigms often treat emotional understanding and generation as separate tasks, yet they inherently possess symbiotic potential for mutual enhancement. In this paper, we aim to bridge the gap by developing a unified framework. The primary challenge lies in the extraction of precise and semantically rich representations of abstract emotions, which are crucial for both tasks. To address this, we harness the Chain-of-Thought reasoning at the latent space of multimodal large language models and propose EmoSym, a unified framework built upon this advanced foundation. Our framework is executed through three key steps: 1) Emotional reasoning knowledge compression. To enable efficient transfer of emotional reasoning priors, we design specialized reasoning tokens to compact emotion-aware contexts from external reasoning knowledge bases into latent representations. 2) Verifiable reinforcement reasoning optimization. To ensure more reliable and consistent emotional reasoning, we develop a verifiable reinforcement learning paradigm to further enhance the reasoning token by emotion-specific verifiable reward signals. Processed through the above two steps, the reasoning token simultaneously enhances emotional understanding while enriching semantic representations, benefiting subsequent emotional generation tasks. 3) Reasoning-augmented generation and online feedback. We then fuse it with emotional representations and feed them into a diffusion model to generate emotion-evoking images. Additionally, to create a generative-to-understanding enhancement feedback, we propose an Online Emotional Memory Bank (OEMB). It leverages newly generated images to progressively update the training dataset in the training process to reinforce understanding. Extensive experiments demonstrate the superior capabilities of our framework in both emotional understanding and generation tasks.

Recent Vision-Language-Action (VLA) models built on pre-trained Vision-Language Models (VLMs) require extensive post-training, resulting in high computational overhead that limits scalability and deployment. Existing sparsification strategies—such as Mixture-of-Depths, layer skipping, and early exit—fall short by neglecting the semantic coupling across vision-language-action modalities, and focusing narrowly on intra-LLM computation while overlooking end-to-end coherence from perception to control. To address these challenges, we propose **CogVLA**, a Cognition-Aligned Vision-Language-Action framework that leverages instruction-driven routing and sparsification to improve both efficiency and performance. CogVLA draws inspiration from human multimodal coordination and introduces a 3-stage progressive architecture. 1) **Encoder-FiLM based Aggregation Routing (EFA-Routing)** injects instruction information into the vision encoder to selectively aggregate and compress dual-stream visual tokens, forming a instruction-aware latent representation. 2) Building upon this compact visual encoding, **LLM-FiLM based Pruning Routing (LFP-Routing)** introduces action intent into the language model by pruning instruction-irrelevant visually grounded tokens, thereby achieving token-level sparsity. 3) To ensure that compressed perception inputs can still support accurate and coherent action generation, we introduce **V‑L‑A Coupled Attention (CAtten)**, which combines causal vision-language attention with bidirectional action parallel decoding. Extensive experiments on the LIBERO benchmark and real-world robotic tasks demonstrate that CogVLA achieves state-of-the-art performance with success rates of 97.4\% and 70.0\%, respectively, while reducing training costs by 2.5$\times$ and decreasing inference latency by 2.8$\times$ compared to OpenVLA.

Multimodal Large Language Models (MLLMs) have gained significant attention due to their rich internal implicit knowledge for cross-modal learning. Although advances in bringing audio-visuals into LLMs have resulted in boosts for a variety of Audio-Visual Question Answering (AVQA) tasks, they still face two crucial challenges: 1) audio-visual ambiguity, and 2) audio-visual hallucination. Existing MLLMs can respond to audio-visual content, yet sometimes fail to describe specific objects due to the ambiguity or hallucination of responses. To overcome the two aforementioned issues, we introduce the CAT+, which enhances MLLM to ensure more robust multimodal understanding. We first propose the Sequential Question-guided Module (SQM), which combines tiny transformer layers and cascades Q-Formers to realize a solid audio-visual grounding. After feature alignment and high-quality instruction tuning, we introduce Ambiguity Scoring Direct Preference Optimization (AS-DPO) to correct the problem of CAT+ bias toward ambiguous descriptions. To explore the hallucinatory deficits of MLLMs in dynamic audio-visual scenes, we build a new Audio-visual Hallucination Benchmark, named AVHbench. This benchmark detects the extent of MLLM's hallucinations across three different protocols in the perceptual object, counting, and holistic description tasks. Extensive experiments across video-based understanding, open-ended, and close-ended AVQA demonstrate the superior performance of our method. The AVHbench is released at https://github.com/rikeilong/Bay-CAT.

Multimodal Large Language Models (MLLMs) have endowed LLMs with the ability to perceive and understand multimodal signals. However, most of the existing MLLMs mainly adopt vision encoders pretrained on coarsely aligned image-text pairs, leading to insufficient extraction and reasoning of visual knowledge. To address this issue, we devise a dual-Level vIsual knOwledge eNhanced Multimodal Large Language Model (LION), which empowers the MLLM by injecting visual knowledge in two levels. 1) Progressive incorporation of fine-grained spatial-aware visual knowledge. We design a vision aggregator cooperated with region-level vision-language (VL) tasks to incorporate fine-grained spatial-aware visual knowledge into the MLLM. To alleviate the conflict between imagelevel and region-level VL tasks during incorporation, we devise a dedicated stage-wise instruction-tuning strategy with mixture-of-adapters. This progressive incorporation scheme contributes to the mutual promotion between these two kinds of VL tasks. 2) Soft prompting of high-level semantic visual evidence. We facilitate the MLLM with high-level semantic visual evidence by leveraging diverse image tags. To mitigate the potential influence caused by imper-fect predicted tags, we propose a soft prompting method by embedding a learnable token into the tailored text instruction. Comprehensive experiments on several multimodal benchmarks demonstrate the superiority of our model (e.g., improvement of 5% accuracy on VSR and 3% CIDEr on TextCaps over InstructBLIP, 5% accuracy on RefCOCOg over Kosmos-2).

Multimodal Large Language Models (MLLMs) have shown impressive reasoning abilities and general intelligence in various domains. It inspires researchers to train end-to-end MLLMs or utilize large models to generate policies with human-selected prompts for embodied agents. However, these methods exhibit limited generalization capabilities on unseen tasks or scenarios, and overlook the multimodal environment information which is critical for robots to make decisions. In this paper, we introduce a novel **Robo**tic **M**ultimodal **P**erception-**P**lanning (**RoboMP$^2$**) framework for robotic manipulation which consists of a Goal-Conditioned Multimodal Preceptor (GCMP) and a Retrieval-Augmented Multimodal Planner (RAMP). Specially, GCMP captures environment states by employing a tailored MLLMs for embodied agents with the abilities of semantic reasoning and localization. RAMP utilizes coarse-to-fine retrieval method to find the $k$ most-relevant policies as in-context demonstrations to enhance the planner. Extensive experiments demonstrate the superiority of RoboMP$^2$ on both VIMA benchmark and real-world tasks, with around 10% improvement over the baselines.

Building a general-purpose agent is a long-standing vision in the field of artificial intelligence. Existing agents have made remarkable progress in many domains, yet they still struggle to complete long-horizon tasks in an open world. We attribute this to the lack of necessary world knowledge and multimodal experience that can guide agents through a variety of long-horizon tasks. In this paper, we propose a Hybrid Multimodal Memory module to address the above challenges. It 1) transforms knowledge into Hierarchical Directed Knowledge Graph that allows agents to explicitly represent and learn world knowledge, and 2) summarises historical information into Abstracted Multimodal Experience Pool that provide agents with rich references for in-context learning. On top of the Hybrid Multimodal Memory module, a multimodal agent, Optimus-1, is constructed with dedicated Knowledge-guided Planner and Experience-Driven Reflector, contributing to a better planning and reflection in the face of long-horizon tasks in Minecraft. Extensive experimental results show that Optimus-1 significantly outperforms all existing agents on challenging long-horizon task benchmarks, and exhibits near human-level performance on many tasks. In addition, we introduce various Multimodal Large Language Models (MLLMs) as the backbone of Optimus-1. Experimental results show that Optimus-1 exhibits strong generalization with the help of the Hybrid Multimodal Memory module, outperforming the GPT-4V baseline on many tasks.

Multimodal large language models (MLLMs) have demonstrated impressive capabilities across various vision-language tasks. However, a generalist MLLM typically underperforms compared with a specialist MLLM on most VL tasks, which can be attributed to task interference. In this paper, we propose a mixture of multimodal experts (MoME) to mitigate task interference and obtain a generalist MLLM. Our MoME is composed of two key components, a mixture of vision experts (MoVE) and a mixture of language experts (MoLE). MoVE can adaptively modulate the features transformed from various vision encoders, and has a strong compatibility in transformation architecture. MoLE incorporates sparsely gated experts into LLMs to achieve painless improvements with roughly unchanged inference costs. In response to task interference, our MoME specializes in both vision and language modality to adapt to task discrepancies. Extensive experiments show that MoME significantly improves the performance of generalist MLLMs across various VL tasks.

Misinformation has become a pressing issue. Fake media, in both visual and textual forms, is widespread on the web. Whilevarious deepfake detection and text fake news detection methods have been proposed, they are only designed for single-modality forgery based on binary classification, let alone analyzing and reasoning subtle forgery traces across different modalities. In this paper, we highlight a new research problem for multi-modal fake media, namelyDetectingandGroundingMulti-ModalMediaManipulation (DGM4).DGM4aims to not only detect the authenticity of multi-modal media, but also ground the manipulated content, which requires deeper reasoning of multi-modal media manipulation. To support a large-scale investigation, we construct the first DGM4 dataset. Moreover, we propose a novelHierArchicalMulti-modalManipulation rEasoning tRansformer (HAMMER) to fully capture the fine-grained interaction between different modalities. HAMMER performs 1) manipulation-aware contrastive learning between two uni-modal encoders as shallow manipulation reasoning, and 2) modality-aware cross-attention by multi-modal aggregator as deep manipulation reasoning. Dedicated manipulation detection and grounding heads are integrated from shallow to deep levels based on the interacted multi-modal information. To exploit more fine-grained contrastive learning for cross-modal semantic alignment, we further integrate Manipulation-Aware Contrastive Loss with Local View and construct a more advanced modelHAMMER++Finally, we build an extensive benchmark and set up rigorous evaluation metrics for this new research problem. Comprehensive experiments demonstrate the superiority ofHAMMERandHAMMER++; several valuable observations are also revealed to facilitate future research in multi-modal media manipulation..

Face presentation attack detection (fPAD) plays a critical role in the modern face recognition pipeline. An fPAD model with good generalization can be obtained when it is trained with face images from different input distributions and different types of spoof attacks. In reality, training data (both real face images and spoof images) are not directly shared between data owners due to legal and privacy issues. In this article, with the motivation of circumventing this challenge, we propose a federated face presentation attack detection (FedPAD) framework that simultaneously takes advantage of rich fPAD information available at different data owners while preserving data privacy. In the proposed framework, each data owner (referred to as data centers) locally trains its own fPAD model. A server learns a global fPAD model by iteratively aggregating model updates from all data centers without accessing private data in each of them. Once the learned global model converges, it is used for fPAD inference. To equip the aggregated fPAD model in the server with better generalization ability to unseen attacks from users, following the basic idea of FedPAD, we further propose a federated generalized face presentation attack detection (FedGPAD) framework. A federated domain disentanglement strategy is introduced in FedGPAD, which treats each data center as one domain and decomposes the fPAD model into domain-invariant and domain-specific parts in each data center. Two parts disentangle the domain-invariant and domain-specific features from images in each local data center. A server learns a global fPAD model by only aggregating domain-invariant parts of the fPAD models from data centers, and thus, a more generalized fPAD model can be aggregated in server. We introduce the experimental setting to evaluate the proposed FedPAD and FedGPAD frameworks and carry out extensive experiments to provide various insights about federated learning for fPAD.

Misinformation has become a pressing issue. Fake media, in both visual and textual forms, is widespread on the web. While various deepfake detection and text fake news detection methods have been proposed, they are only designed for single-modality forgery based on binary classification, let alone analyzing and reasoning subtle forgery traces across different modalities. In this paper, we high-light a new research problem for multi-modal fake media, namely Detecting and Grounding Multi-Modal Media Manipulation (DGM4). DGM4aims to not only detect the authenticity of multi-modal media, but also ground the manipulated content (i.e., image bounding boxes and text tokens), which requires deeper reasoning of multi-modal media manipulation. To support a large-scale investigation, we construct the first DGM4dataset, where image-text pairs are manipulated by various approaches, with rich annotation of diverse manipulations. Moreover, we propose a novel HierArchical Multi-modal Manipulation rEasoning tRansformer (HAMMER) to fully capture the fine-grained interaction between different modalities. HAMMER performs 1) manipulation-aware contrastive learning between two uni-modal encoders as shallow manipulation reasoning, and 2) modality-aware cross-attention by multi-modal aggregator as deep manipulation reasoning. Dedicated manipulation detection and grounding heads are integrated from shallow to deep levels based on the interacted multi-modal information. Finally, we build an extensive bench-mark and set up rigorous evaluation metrics for this new research problem. Comprehensive experiments demonstrate the superiority of our model; several valuable observations are also revealed to facilitate future research in multi-modal media manipulation.

Test-time training (TTT) is an emerging approach for addressing the problem of domain shift. In its framework, a test-time training phase is inserted between the training phase and the test phase. During the test-time training phase, the representation layers are adapted using an auxiliary task. Then the updated model will be used in the test phase. Although the idea is very intuitive, TTT does not demonstrate competitive performance compared with some other domain adaption methods. In this paper, we present both theoretical and empirical analyses to explain the subpar performance of TTT. In particular, we point out that TTT causes a new kind of problem, which we term as Model Mismatch. To address this problem of Model Mismatch, we analyse a simple yet effective method inspired by the idea of mixup in robust training. Such effectiveness is shown in the experimental results.

Face presentation attack detection (fPAD) plays a critical role in the modern face recognition pipeline. The generalization ability of face presentation attack detection models to unseen attacks has become a key issue for real-world deployment, which can be improved when models are trained with face images from different input distributions and different types of spoof attacks. In reality, due to legal and privacy issues, training data (both real face images and spoof images) are not allowed to be directly shared between different data sources. In this paper, to circumvent this challenge, we propose a Federated Test-Time Adaptive Face Presentation Attack Detection with Dual-Phase Privacy Preservation framework, with the aim of enhancing the generalization ability of fPAD models in both training and testing phase while preserving data privacy. In the training phase, the proposed framework exploits the federated learning technique, which simultaneously takes advantage of rich fPAD information available at different data sources by aggregating model updates from them without accessing their private data. To further boost the generalization ability, in the testing phase, we explore test-time adaptation by minimizing the entropy of fPAD model prediction on the testing data, which alleviates the domain gap between training and testing data and thus reduces the generalization error of a fPAD model. We introduce the experimental setting to evaluate the proposed framework and carry out extensive experiments to provide various insights about the proposed method for fPAD.

Face presentation attacks have become an increasingly critical concern when face recognition is widely applied. Many face anti-spoofing methods have been proposed, but most of them ignore the generalization ability to unseen attacks. To overcome the limitation, this work casts face anti-spoofing as a domain generalization (DG) problem, and attempts to address this problem by developing a new meta-learning framework called Regularized Fine-grained Meta-learning. To let our face anti-spoofing model generalize well to unseen attacks, the proposed framework trains our model to perform well in the simulated domain shift scenarios, which is achieved by finding generalized learning directions in the meta-learning process. Specifically, the proposed framework incorporates the domain knowledge of face anti-spoofing as the regularization so that meta-learning is conducted in the feature space regularized by the supervision of domain knowledge. This enables our model more likely to find generalized learning directions with the regularized meta-learning for face anti-spoofing task. Besides, to further enhance the generalization ability of our model, the proposed framework adopts a fine-grained learning strategy that simultaneously conducts meta-learning in a variety of domain shift scenarios in each iteration. Extensive experiments on four public datasets validate the effectiveness of the proposed method.

Face presentation attacks have become an increasingly critical issue in the face recognition community. Many face anti-spoofing methods have been proposed, but they cannot generalize well on "unseen" attacks. This work focuses on improving the generalization ability of face anti-spoofing methods from the perspective of the domain generalization. We propose to learn a generalized feature space via a novel multi-adversarial discriminative deep domain generalization framework. In this framework, a multi-adversarial deep domain generalization is performed under a dual-force triplet-mining constraint. This ensures that the learned feature space is discriminative and shared by multiple source domains, and thus is more generalized to new face presentation attacks. An auxiliary face depth supervision is incorporated to further enhance the generalization ability. Extensive experiments on four public datasets validate the effectiveness of the proposed method.

Unsupervised visual domain adaptation aims to train a classifier that works well on a target domain given labelled source samples and unlabelled target samples. The key issue in unsupervised visual domain adaptation is how to do the feature alignment between source and target domains. Inspired by the adversarial learning in generative adversarial networks, this study proposes a novel adversarial auto‐encoder for unsupervised deep domain adaptation. This method incorporates the auto‐encoder with the adversarial learning so that the domain similarity and reconstruction information from the decoder can be exploited to facilitate the adversarial domain adaptation in the encoder. Extensive experiments on various visual recognition tasks show that the proposed method performs favourably against or better than competitive state‐of‐the‐art methods.

Three-dimensional mask spoofing attacks have been one of the main challenges in face recognition. Compared with a 3D mask, a real face displays different facial motion patterns that are reflected by different facial dynamic textures. However, a large portion of these facial motion differences is subtle. We find that the subtle facial motion can be fully captured by multiple deep dynamic textures from a convolutional layer of a convolutional neural network, but not all deep dynamic textures from different spatial regions and different channels of a convolutional layer are useful for differentiation of subtle motions between real faces and 3D masks. In this paper, we propose a novel feature learning model to learn discriminative deep dynamic textures for 3D mask face anti-spoofing. A novel joint discriminative learning strategy is further incorporated in the learning model to jointly learn the spatial- and channel-discriminability of the deep dynamic textures. The proposed joint discriminative learning strategy can be used to adaptively weight the discriminability of the learned feature from different spatial regions or channels, which ensures that more discriminative deep dynamic textures play more important roles in face/mask classification. Experiments on several publicly available data sets validate that the proposed method achieves promising results in intra- and cross-data set scenarios.

In multimedia analysis, one objective of unsupervised visual domain adaptation is to train a classifier that works well on a target domain given labeled source samples and unlabeled target samples. Feature alignment of two domains is the key issue which should be addressed to achieve this objective. Inspired by the recent study of Generative Adversarial Networks (GAN) in domain adaptation, this paper proposes a new model based on Generative Adversarial Network, named Hierarchical Adversarial Deep Network (HADN), which jointly optimizes the feature-level and pixel-level adversarial adaptation within a hierarchical network structure. Specifically, the hierarchical network structure ensures that the knowledge from pixel-level adversarial adaptation can be back propagated to facilitate the feature-level adaptation, which achieves a better feature alignment under the constraint of pixel-level adversarial adaptation. Extensive experiments on various visual recognition tasks show that the proposed method performs favorably against or better than competitive state-of-the-art methods.

In recent years, software-based face presentation attack detection (PAD) methods have seen a great progress. However, most existing schemes are not able to generalize well in more realistic conditions. The objective of this competition is to evaluate and compare the generalization performances of mobile face PAD techniques under some real-world variations, including unseen input sensors, presentation attack instruments (PAI) and illumination conditions, on a larger scale OULU-NPU dataset using its standard evaluation protocols and metrics. Thirteen teams from academic and industrial institutions across the world participated in this competition. This time typical liveness detection based on physiological signs of life was totally discarded. Instead, every submitted system relies practically on some sort of feature representation extracted from the face and/or background regions using hand-crafted, learned or hybrid descriptors. Interesting results and findings are presented and discussed in this paper.

3D mask spoofing attack has been one of the main challenges in face recognition. A real face displays a different motion behaviour compared to a 3D mask spoof attempt, which is reflected by different facial dynamic textures. However, the different dynamic information usually exists in the subtle texture level, which cannot be fully differentiated by traditional hand-crafted texture-based methods. In this paper, we propose a novel method for 3D mask face anti-spoofing, namely deep convolutional dynamic texture learning, which learns robust dynamic texture information from fine-grained deep convolutional features. Moreover, channel-discriminability constraint is adaptively incorporated to weight the discriminability of feature channels in the learning process. Experiments on both public datasets validate that the proposed method achieves promising results under intra and cross dataset scenario.