论文检索

Shilin Zhang, Suixue Wang, Qingchen Zhang, Xiulong Liu

Recent advances in spatial transcriptomics have enabled the integration of gene expression profiles with precise spatial coordinates, which have facilitated the exploration of tumor occurrence and development mechanisms, as well as the development of more effective targeted and immunotherapy approaches for tumor treatment. Deciphering cell type represents a critical challenge in spatial transcriptomics research. Existing methods are limited by the pervasive “dropout” events in spatial transcriptomics, hindering their ability to fully capture the relationship between spatial location and gene expression, thereby compromising the performance of cell type deconvolution. To address these limitations, we propose a spatial-aware masked graph transformer-diffusion model (SAMGTD) for enhanced cell type deconvolution in spatial transcriptomics. For spatial transcriptomics, the masked graph transformer model is designed to adaptively capture complex dependencies between spatial locations and gene expression. It employs a masking strategy that guides the model to focus on important local information during training, while the multi-head attention mechanism captures global context. More importantly, the spatial diffusion model is constructed to achieve the dual enhancement of spatial transcriptomics, including denoising and data imputation. It incorporates the multi-head attention mechanism and residual blocks, effectively addressing the “dropout” issue commonly encountered in spatial transcriptomics. For scRNA-seq, we construct a variational autoencoder to reduce noise interference while preserving key gene expression information. Finally, we construct a spatial-aware contrastive learning model to integrate scRNA-seq and spatial transcriptomics for cell type deconvolution. Experiments conducted on three datasets demonstrate that SAMGTD outperforms baseline methods.

Shilin Zhang, Weiliang Huo, Qingchen Zhang, Xiulong Liu

Recent advances in spatial transcriptomics have enabled the simultaneous measurement of gene expression profiles and spatial location information, offering a more comprehensive and in-depth view for studying the tissue microenvironment. Spatial domain identification is a crucial step in analyzing spatial transcriptomics. However, current methods have poor accuracy and visualization because they lack self-adaptability to different tissue data, and moreover, they cannot effectively extract spatial location information. To address these issues, we propose an adaptive graph contrastive learning framework based on multi-head graph attention networks (GATCL) for spatial domain identification. Specifically, we design a data augmentation module to mask and shuffle the pre-processed gene expression data to generate more differentiated negative samples. In addition, we construct the multi-head graph attention networks (MHGAT) to encode gene expression profiles and spatial location information. More importantly, we design an adaptive graph contrastive learning model that works both with positive and negative samples from spatial transcriptomics. We introduce the attention pooling mechanism to dynamically and adaptively aggregate the spots' neighborhood information, and to improve the model's generalization ability for different spatial transcriptomics data. Furthermore, we design a discriminator that adds spectral normalization to bilinear functions. Experimental results on DLPFC, breast cancer, and mouse somatosensory cortex datasets demonstrate that the average Adjusted Rand Index (ARI) scores are 0.5746, 0.6182, and 0.6496, respectively, significantly outperforming baseline methods. More importantly, GATCL provides a more detailed visualization of different spatial transcriptomics data.

Lin Zhang, Shikui Tu, Lei Xu

Electroencephalography (EEG) plays a vital role in clinical and cognitive applications such as epilepsy diagnosis and emotion recognition. However, the low signal-to-noise ratio, inter-subject variability, and inherent non-stationarity of EEG signals present substantial modeling challenges. While recent Transformer-based models offer promising long-range modeling capabilities, their self-attention mechanism behaves as a low-pass filter, suppressing high-frequency neural patterns critical for decoding transient events. In this work, we provide the first formal analysis demonstrating this low-pass behavior in self-attention mechanisms when applied to EEG signals, revealing a fundamental limitation of deep attention-based EEG models. To address this, we propose SEBSFormer, a spectral-enhanced bi-Stream Transformer that jointly models temporal dependencies and spectral structures. SEBSFormer integrates three key modules: a spectral compensation module that restores high-frequency components via residual correction in the Fourier domain; a multi-scale temporal attention module for saliency-guided temporal compression; and a graph-guided dynamic fusion module for adaptive spatial aggregation across electrodes. Extensive experiments on three benchmark datasets—TUAB, TUEV, and SEED—demonstrate that SEBSFormer consistently outperforms existing state-of-the-art models across both clinical and affective tasks. Our findings establish a new paradigm for frequency-aware EEG modeling.

Jiaqi Zhang, Chen Gao, Liyuan Zhang, Quoc Viet Hung Nguyen, Hongzhi Yin

Recent advances in embodied agents with multimodal perception and reasoning capabilities based on large vision-language models (LVLMs), excel in autonomously interacting either real or cyber worlds, helping people make intelligent decisions in complex environments. However, the current works are normally optimized by golden action trajectories or ideal task-oriented solutions toward a definitive goal. This paradigm considers limited user-oriented factors, which could be the reason for their performance reduction in a wide range of personal assistant applications. To address this, we propose Chain-of-User-Thought (COUT, a novel embodied reasoning paradigm that takes a chain of thought from basic action thinking to explicit and implicit personalized preference thought to incorporate personalized factors into autonomous agent learning. The main challenges of achieving COUT include: 1) the definition of embodied personalized tasks, 2) the embodied environment epitomizes personalized preference, and 3) the way to model embodied personalized actions. To target COUT, we introduce SmartAgent, an agent framework perceiving cyber environments and reasoning personalized requirements as: 1) interacting with GUI to access an item pool, 2) generating users' explicit requirements implied by previous actions, and 3) recommending items to fulfill users' implicit requirements. To demonstrate SmartAgent's capabilities, we also create a brand-new dataset SmartSpot that offers a full-stage personalized action-involved environment. To our best knowledge, our work is the first to formulate the COUT process, serving as a preliminary attempt towards embodied personalized agent learning. Our extensive experiments on SmartSpot illuminate SmartAgent’s functionality among a series of embodied and personalized sub-tasks.

Jiayue Yuan, Fangting Xie, Guangwen Ouyang, Changhai Ma, Ziyu Wu, Heyu Ding, Quan Wan, Yi Ke, Yuchen Wu, Xiaohui Cai

Multi-person global human mesh recovery (HMR) is crucial for understanding crowd dynamics and interactions. Traditional vision-based HMR methods sometimes face limitations in real-world scenarios due to mutual occlusions, insufficient lighting, and privacy concerns. Human-floor tactile interactions offer an occlusion-free and privacy-friendly alternative for capturing human motion. Existing research indicates that pressure signals acquired from tactile mats can effectively estimate human pose in single-person scenarios. However, when multiple individuals walk randomly on the mat simultaneously, how to distinguish intermingled pressure signals generated by different persons and subsequently acquire individual temporal pressure data remains a pending challenge for extending pressure-based HMR to the multi-person situation. In this paper, we present PressTrack-HMR, a top-down pipeline that recovers multi-person global human meshes solely from pressure signals. This pipeline leverages a tracking-by-detection strategy to first identify and segment each individual's pressure signal from the raw pressure data, and subsequently performs HMR for each extracted individual signal. Furthermore, we build a multi-person interaction pressure dataset MIP, which facilitates further research into pressure-based human motion analysis in multi-person scenarios. Experimental results demonstrate that our method excels in multi-person HMR using pressure data, with 89.2 MPJPE and 112.6 WA-MPJPE, and these showcase the potential of tactile mats for ubiquitous, privacy-preserving multi-person action recognition.

Yan Yu, Yilun Liu, Minggui He, Shimin Tao, Weibin Meng, Xinhua Yang, Li Zhang, Hongxia Ma, Dengye Li, Daimeng Wei 等

Pairwise evaluation of large language models (LLMs) has become the dominant paradigm for benchmarking open-ended tasks, yet non-transitive preferences—where evaluators prefer A over B, B over C, but C over A—fundamentally undermine ranking reliability. We show that this critical issue stems largely from low-quality data that contains inherently ambiguous preference pairs. To address this challenge, we propose ELSPR, a principled graph-theoretic framework that models pairwise preferences as tournament graphs and systematically identifies problematic training data. ELSPR quantifies non-transitivity through strongly connected components (SCCs) analysis and measures overall preference clarity using a novel normalized directed graph structural entropy metric. Our filtering methodology selectively removes preference data that induce non-transitivity while preserving transitive preferences. Extensive experiments on the AlpacaEval benchmark demonstrate that models fine-tuned on ELSPR-filtered data achieve substantial improvements: a 13.8% reduction in non-transitivity, a 0.088 decrease in structural entropy, and significantly enhanced discriminative power in real-world evaluation systems. Human validation confirms that discarded data exhibit dramatically lower inter-annotator agreement (34.4% vs. 52.6%) and model-human consistency (51.2% vs. 80.6%) compared to cleaned data. These findings establish ELSPR as an effective data self-purification approach for developing more robust, consistent, and human-aligned LLM evaluation systems.

Xinyue Yu, Youqing Fang, Pingyu Wu, Guoyang Ye, Wenbo Zhou, Weiming Zhang, Song Xiao

Generating expressive and controllable human speech is one of the core goals of generative artificial intelligence, but its progress has long been constrained by two fundamental challenges: the deep entanglement of speech factors and the coarse granularity of existing control mechanisms. To overcome these challenges, we have proposed a novel framework called MF-Speech, which consists of two core components: MF-SpeechEncoder and MF-SpeechGenerator. MF-SpeechEncoder acts as a factor purifier, adopting a multi-objective optimization strategy to decompose the original speech signal into highly pure and independent representations of content, timbre, and emotion. Subsequently, MF-SpeechGenerator functions as a conductor, achieving precise, composable and fine-grained control over these factors through dynamic fusion and Hierarchical Style Adaptive Normalization (HSAN). Experiments demonstrate that in the highly challenging multi-factor compositional speech generation task, MF-Speech significantly outperforms current state-of-the-art methods, achieving a lower word error rate (WER=4.67%), superior style control (SECS=0.5685, Corr=0.68), and the highest subjective evaluation scores (nMOS=3.96, sMOS_t=3.86, sMOS_e=3.78). Furthermore, the learned discrete factors exhibit strong transferability, demonstrating their significant potential as a general-purpose speech representation.

Longlong Yu, Wenxi Li, Yaoqi Sun, Hang Xu, Chenggang Yan, Yuchen Guo

Despite recent progress in adapting State Space Models such as Mamba to vision tasks, their intrinsic 1D scanning mechanism imposes limitations when applied to inherently 2D-structured data like images. Existing adaptations, including VMamba and 2DMamba, either suffer from inconsistency between scanning order and spatial locality or restrict inter-patch communication to singular paths, hindering effective information propagation. In this paper, we propose 2D-CrossScan, a novel 2D-compatible scan framework that enables spatially consistent, multi-path hidden state propagation by integrating modified state equations over two-dimensional neighborhoods. Furthermore, we mitigate redundant information accumulation due to overlapping paths via cross-directional subtraction. To fully align with the 2D spatial structure, we introduce a multi-directional scanning strategy that starts simultaneously from all four corners of the image, enabling diverse propagation paths and better feature integration. Our approach maintains efficiency, requiring only minimal architectural changes to existing Mamba variants. Experimental results demonstrate substantial improvements in multiple visual tasks, including object detection and semantic segmentation on PANDA and COCO datasets. Compared to baseline SSM-based methods, 2D-CrossScan consistently yields better spatial representations, as confirmed by extensive effective receptive field visualizations and attention analyses. These results highlight the importance of geometry-aware state propagation and validate 2D-CrossScan as a simple yet powerful extension to SSMs for vision.

Wen Yin, Siyu Zhan, Cencen Liu, Xin Hu, Guiduo Duan, Xiurui Xie, Yuan-Fang Li, Tao He

Multimodal Emotion Recognition (MER) aims to accurately identify human emotional states by integrating heterogeneous modalities such as visual, auditory, and textual data. Existing approaches predominantly rely on unified emotion labels to supervise model training, often overlooking a critical challenge: inter-modal emotion conflicts, wherein different modalities within the same sample may express divergent emotional tendencies. In this work, we address this overlooked issue by proposing a novel framework, Typicality-based Consistent-aware Multimodal Emotion Recognition (TiCAL), inspired by the stage-wise nature of human emotion perception. TiCAL dynamically assesses the consistency of each training sample by leveraging pseudo unimodal emotion labels alongside a typicality estimation. To further enhance emotion representation, we embed features in a hyperbolic space, enabling the capture of fine-grained distinctions among emotional categories. By incorporating consistency estimates into the learning process, our method improves model performance, particularly on samples exhibiting high modality inconsistency. Extensive experiments on benchmark datasets, e.g, MOSEI and MER2023, validate the effectiveness of TiCAL in mitigating inter-modal emotional conflicts and enhancing overall recognition accuracy, e.g., with about 2.6% improvements over the state-of-the-art DMD.

Zipeng Ye, Wenjian Luo, Qi Zhou, Yubo Tang

Large language models (LLMs) have seen remarkable growth in recent years. To leverage convenient LLM cloud services, users are inevitably to upload their prompts. Additionally, for tasks such as translation, reading comprehension, and summarization, associated files or context are inherently needed, whether or not they contain user privacy information. Despite the rapid progress in LLM capabilities, research on preserving user privacy during inference has been relatively scarce. To this end, this paper conducts some exploratory research in this domain. Firstly, we show that (1) the embedding space of tokens is highly sparse, and (2) LLMs primarily function in the orthogonal subspace of embedding space, these two factors making privacy extremely vulnerable. Then, we analyze the structural characteristics of LLMs and design a distributed privacy-preserving inference paradigm which can effectively resist privacy attacks. Finally, we perform a thorough evaluation of the defended models on mainstream tasks and find that low-bit quantization techniques can be effectively combined with our inference paradigm, achieving a balance between privacy, utility, and runtime memory efficiency.

Qilang Ye, Yu Zhou, Lian He, Jie Zhang, Xuanming Guo, Jiayu Zhang, Mingkui Tan, Weicheng Xie, Yue Sun, Tao Tan 等

Large Language Models (LLMs) hold rich implicit knowledge and powerful transferability. In this paper, we explore the combination of LLMs with the human skeleton to perform action classification and description. However, when treating LLM as a recognizer, two questions arise: 1) How can LLMs understand the skeleton? 2) How can LLMs distinguish among actions? To address these problems, we introduce a novel paradigm named learning Skeleton representation with visual-motion knowledge for Action Recognition (SUGAR). In our pipeline, we first utilize off-the-shelf large-scale video models as a knowledge base to generate visual, motion information related to actions. Then, we propose to supervise skeleton learning through this prior knowledge to yield discrete representations. Finally, we use the LLM with untouched pre-training weights to understand these representations and generate the desired action targets and descriptions. Notably, we present a Temporal Query Projection (TQP) module to continuously model the skeleton signals with long sequences. Experiments on several skeleton-based action classification benchmarks demonstrate the efficacy of our SUGAR. Moreover, experiments on zero-shot scenarios show that SUGAR is more versatile than linear-based methods.

Sicheng Yang, Yukai Huang, Weitong Cai, Shitong Sun, You He, Jiankang Deng, Hang Zhang, Jifei Song, Zhensong Zhang

The performance of egocentric AI agents is fundamentally limited by multimodal intent ambiguity. This challenge arises from a combination of underspecified language, imperfect visual data, and deictic gestures, which frequently leads to task failure. Existing monolithic Vision-Language Models (VLMs) struggle to resolve these multimodal ambiguous inputs, often failing silently or hallucinating responses. To address these ambiguities, we introduce the Plug-and-Play Clarifier, a zero-shot and modular framework that decomposes the problem into discrete, solvable sub-tasks. Specifically, our framework consists of three synergistic modules: (1) a text clarifier that uses dialogue-driven reasoning to interactively disambiguate linguistic intent, (2) a vision clarifier that delivers real-time guidance feedback, instructing users to adjust their positioning for improved capture quality, and (3) a cross-modal clarifier with grounding mechanism that robustly interprets 3D pointing gestures and identifies the specific objects users are pointing to. Extensive experiments demonstrate that our framework improves the intent clarification performance of small language models (4-8B) by approximately 30%, making them competitive with significantly larger counterparts. We also observe consistent gains when applying our framework to these larger models. Furthermore, our vision clarifier increases corrective guidance accuracy by over 20%, and our cross-modal clarifier improves semantic answer accuracy for referential grounding by 5%. Overall, our method provides a plug-and-play framework that effectively resolves multimodal ambiguity and significantly enhances user experience in egocentric interaction.

Hang Yang, Zhiwu Li, Witold Pedrycz

Crowdsourcing is a common approach for training data-hungry models by collecting high-quality labeled data with human labor. With crowdsourcing data, the end-to-end learning paradigm is rising, where the classifier is concatenated with annotator-specific confusion layers and the two parts are co-trained in a parameter-coupled manner. However, learning with the size of a very large set of annotations is a challenge when computation or energy is limited. In this paper, we analyze and refine the coresets for end-to-end learning from crowds under the sensitivity sampling framework. This coreset is a small possible subset of annotations, so one can efficiently optimize the Coupled Cross-Entropy Minimization problem with guaranteed approximation. We first prove the lower bound, which shows no coresets smaller than complete data with confusion layers. Then, with workers' transition matrices, we show that with the regularization term, this lower bound can be prevented. Our main result is that under mild assumptions, a smaller coreset exists for the regularized Coupled Cross-Entropy Minimization problem. An upper bound of sensitivity is proposed for designing a sampling algorithm called CrowdCore. The experimental results on synthetic and real-world datasets demonstrate the effectiveness of our analysis.

Liqi Yan, Yihao Wu, Chenyi Xu, Chao Yang, Jianhui Zhang, Pan Li

Augmented Reality (AR) navigation has emerged as a transformative tool for spatial intelligence, enabling users to interactively explore complex environments through wearable and mobile AR devices. However, current AR navigation systems struggle with low indoor localization accuracy, weak semantic understanding, and limited long-term memory, which severely limits their adaptability in dynamic, multi-floor, and large-scale real-world settings. To address these challenges, we present AR-Nav benchmark, a novel dataset with corresponding suite that leverages vision and language for AR navigation. First, to construct this benchmark, we proposed an Augmented Reality Visual-Language Memory Model (AR‑VLM²), which generates structured, semantically rich, and temporally indexed representations for long-term AR navigation. Second, we design a lightweight navigation intent recommending module with hierarchical topological reasoning and language-grounded path planning, called ARN‑Pilot, enabling low-latency and personalized route selection. Third, we introduce a closed-loop AR interaction module that supports real-time multi-modal feedback, dynamic memory updates, and human-in-the-loop query refinement. Extensive experiments in indoor multi-floor and outdoor parking scenarios show that AR-Nav suite significantly outperforms state-of-the-art AR navigation methods.

Hongbin Xu, Chaohui Yu, Feng Xiao, Jiazheng Xing, Hai Ci, Weitao Chen, Fan Wang, Ming Li

Despite the remarkable progress of 3D generation, achieving controllability, i.e., ensuring consistency between generated 3D content and input conditions like edge and depth, remains a significant challenge. Existing methods often struggle to maintain accurate alignment, leading to noticeable discrepancies. To address this issue, we propose CyC3D, a new framework that enhances controllable 3D generation by explicitly encouraging cyclic consistency between the second-order 3D content, generated based on extracted signals from the first-order generation, and its original input controls. Specifically, we employ an efficient feed-forward backbone that can generate a 3D object from an input condition and a text prompt. Given an initial viewpoint and a control signal, a novel view is rendered from the generated 3D content, from which the extracted condition is used to regenerate the 3D content. This re-generated output is then rendered back to the initial viewpoint, followed by another round of control signal extraction, forming a cyclic process with two consistency constraints. View consistency ensures coherence between the two generated 3D objects, measured by semantic similarity to accommodate generative diversity. Condition consistency aligns the final extracted signal with the original input control, preserving structural or geometric details throughout the process. Extensive experiments on popular benchmarks demonstrate that CyC3D significantly improves controllability, especially for fine-grained details, outperforming existing methods across various conditions (e.g., +14.17% PSNR for edge, +6.26% PSNR for sketch).

Jian Xiong, Lingxia Jiang, Xianzhong Long, Miaohui Wang, Hao Gao

Point cloud quality assessment (PCQA) is essential for reliable 3D visual applications. While point-based methods face challenges in characterizing distortions due to point cloud disorder, projection-based approaches offer better efficiency but suffer from geometric distortion insensitivity and texture representation blind spots. This study proposes SAF-Net, a multi-view structure-aware feature fusion network for PCQA. We first identify two key limitations in projection-based methods: insufficient geometric distortion perception and representation blind spots (RBS) in texture images. To address these issues, SAF-Net innovatively integrates object mask maps and local binary pattern (LBP) maps. The mask maps enhance geometric distortion perception by extracting edge sharpness and curvature variations, while LBP maps capture essential structural information to overcome RBS and align with human visual system (HVS) sensitivity. SAF-Net employs a hybrid CNN-ViT architecture to balance local feature extraction and global context modeling, along with a progressive fusion strategy to optimize cross-modal feature interaction. Extensive experiments demonstrate the superior performance of SAF-Net on multiple benchmarks, establishing new state-of-the-art results in PCQA.

Xiaohan Xie, Haoran Yu, Biying Shou, Jianwei Huang

What if machines could discover human behavioral patterns better than experts? Traditional behavioral modeling in economics depends on costly manual refinement by domain experts, severely limiting scalability and discovery potential. We introduce AutoBM, an automated behavioral modeling framework leveraging large language models (LLMs) to systematically generate, evaluate, and refine interpretable behavioral models directly from human behavior data. AutoBM represents candidate models as structured natural language specifications, explicitly defining symbolic terms along with their tunable parameters, interpretations, and design rationales. AutoBM leverages LLMs to automatically translate each language specification into executable code, optimize tunable parameters, and evaluate model performance. Utilizing LLM-guided search strategies, AutoBM iteratively recombines and improves models at the term level, closely mirroring human expert practices. Experiments conducted across three distinct strategic environments (the ultimatum game, repeated rock-paper-scissors, and continuous double auctions) demonstrate that AutoBM-generated models consistently outperform leading manually crafted models, achieving significant improvements in prediction accuracy while maintaining clear interpretability. Our results demonstrate that automated frameworks can not only match but systematically exceed human expertise in behavioral modeling, fundamentally changing how we understand strategic human behavior.

Weihao Xia, Cengiz Oztireli

Existing evaluation protocols for brain visual decoding predominantly rely on coarse metrics that obscure inter-model differences, lack neuroscientific foundation, and fail to capture fine-grained visual distinctions. To address these limitations, we introduce BASIC, a unified, multigranular evaluation framework that jointly quantifies structural fidelity, inferential alignment, and contextual coherence between decoded and ground-truth images. For the structural level, we introduce a hierarchical suite of segmentation-based metrics, including foreground, semantic, instance, and component masks, anchored in granularity-aware correspondence across mask structures. For the semantic level, we extract structured scene representations encompassing objects, attributes, and relationships using multimodal large language models, enabling detailed, scalable, and context-rich comparisons with ground-truth stimuli. We benchmark a diverse set of visual decoding methods across multiple stimulus-neuroimaging datasets within this unified evaluation framework. Together, these criteria provide a more discriminative, interpretable, and comprehensive foundation for evaluating brain visual decoding methods.

Lukun Wu, Jie Li, Ziqi Ren, Kaifan Zhang, Xinbo Gao

Decoding visual features from EEG signals is a central challenge in neuroscience, with cross-modal alignment as the dominant approach. We argue that the relationship between visual and brain modalities is fundamentally asymmetric, characterized by two critical gaps: a Fidelity Gap (stemming from EEG's inherent noise and signal degradation, vs. vision's high-fidelity features) and a Semantic Gap (arising from EEG's shallow conceptual representation, vs. vision's rich semantic depth). Previous methods often overlook this asymmetry, forcing alignment between the two modalities as if they were equal partners and thereby leading to poor generalization. To address this, we propose the adaptive teaching paradigm. This paradigm empowers the ``teacher" modality (vision) to dynamically shrink and adjust its knowledge structure under task guidance, tailoring its semantically dense features to match the ``student" modality (EEG)'s capacity. We implement this paradigm with the ShrinkAdapter, a simple yet effective module featuring a residual-free design and a bottleneck structure. Through extensive experiments, we validate the underlying rationale and effectiveness of our paradigm. Our method achieves a top-1 accuracy of 60.2% on the zero-shot brain-to-image retrieval task, surpassing previous state-of-the-art methods by a margin of 9.8%. Our work introduces a new perspective for asymmetric alignment: the teacher must shrink and adapt to bridge the vision-brain gap.

Weining Weng, Yang Gu, Yuan Ma, Yuchen Liu, Yingwei Zhang, Yiqiang Chen

Cross-subject EEG decoding remains a fundamental challenge due to substantial inter-subject variability in brain activity, which hinders the development of subject-independent EEG models. Despite progress in extracting cross-subject invariant features, existing studies neglect the shared neural responses that arise under similar cognitive or emotional states across individuals, limiting their ability to learn generalized and consistent EEG representations. To address the challenges, we propose State Mamba, a novel spatiotemporal EEG state-space model that explicitly models and aligns neural responses and their spatiotemporal state transitions to learn consistent and generalizable representations across subjects. Innovatively, State Mamba theoretically formulates a multi-channel Mamba architecture that jointly models spatial and temporal brain state transitions, supporting principled analysis of neural responses. To enhance spatiotemporal feature coupling, we introduce the LGANN module, which adopts global-local attention to integrate long- and short-term brain activity into a compact EEG representation. Furthermore, we design two self-supervised pretext tasks to extract consistent neural patterns across subjects: (1) representation alignment to align EEG representation, and (2) pattern alignment to align their transition rules under identical conditions, jointly promoting subject-invariant EEG representations. Extensive experiments on three benchmark datasets, FACED, DEAP, and ISRUC, demonstrate the superior performance of State Mamba in cross-subject emotion and sleep recognition tasks, validating its robust generalization capability.