Large-scale Models Research Articles

A general framework for characterizing optimal communication in brain networks

Efficient communication in brain networks is foundational for cognitive function and behavior. However, how communication efficiency is defined depends on the assumed model of signaling dynamics, e.g., shortest path signaling, random walker navigation, broadcasting, and diffusive processes. Thus, a general and model-agnostic framework for characterizing optimal neural communication is needed. We address this challenge by assigning communication efficiency through a virtual multi-site lesioning regime combined with game theory, applied to large-scale models of human brain dynamics. Our framework quantifies the exact influence each node exerts over every other, generating optimal influence maps given the underlying model of neural dynamics. These descriptions reveal how communication patterns unfold if regions are set to maximize their influence over one another. Comparing these maps with a variety of brain communication models showed that optimal communication closely resembles a broadcasting regime in which regions leverage multiple parallel channels for information dissemination. Moreover, we found that the brain’s most influential regions are its rich-club, exploiting their topological vantage point by broadcasting across numerous pathways that enhance their reach even if the underlying connections are weak. Altogether, our work provides a rigorous and versatile framework for characterizing optimal brain communication, and uncovers the most influential brain regions, and the topological features underlying their influence.

Diffusing Winding Gradients (DWG): A Parallel and Scalable Method for 3D Reconstruction from Unoriented Point Clouds

This article presents Diffusing Winding Gradients (DWG) for reconstructing watertight surfaces from unoriented point clouds. Our method exploits the alignment between the gradients of the screened generalized winding number (GWN) field—a robust variant of the standard GWN field—and globally consistent normals to orient points. Starting with an unoriented point cloud, DWG initially assigns a random normal to each point. It computes the corresponding screened GWN field and extracts a level set whose iso-value is the average of GWN values across all input points. The gradients of this level set are then utilized to update the point normals. This cycle of recomputing the screened GWN field and updating point normals is repeated until the screened GWN level sets stabilize and their gradients cease to change. Unlike conventional methods, DWG does not rely on solving linear systems or optimizing objective functions, which simplifies its implementation and enhances its suitability for efficient parallel execution. Experimental results demonstrate that DWG significantly outperforms existing methods in terms of runtime performance. For large-scale models with 10 to 20 million points, our CUDA implementation on an NVIDIA GTX 4090 GPU achieves speeds 30 to 120 times faster than iPSR, the leading sequential method, tested on a high-end PC with an Intel i9 CPU. Furthermore, by employing a screened variant of GWN, DWG demonstrates enhanced robustness against noise and outliers and proves effective for models with thin structures and real-world inputs with overlapping and misaligned scans. For source code and additional results, visit our project webpage: https://dwgtech.github.io/ .

Secure Hierarchical Federated Learning for Large-Scale AI Models: Poisoning Attack Defense and Privacy Preservation in AIoT

The rapid integration of large-scale AI models into distributed systems, such as the Artificial Intelligence of Things (AIoT), has introduced critical security and privacy challenges. While configurable models enhance resource efficiency, their deployment in heterogeneous edge environments remains vulnerable to poisoning attacks, data leakage, and adversarial interference, threatening the integrity of collaborative learning and responsible AI deployment. To address these issues, this paper proposes a Hierarchical Federated Cross-domain Retrieval (FHCR) framework tailored for secure and privacy-preserving AIoT systems. By decoupling models into a shared retrieval layer (globally optimized via federated learning) and device-specific layers (locally personalized), FHCR minimizes communication overhead while enabling dynamic module selection. Crucially, we integrate a retrieval-layer mean inspection (RLMI) mechanism to detect and filter malicious gradient updates, effectively mitigating poisoning attacks and reducing attack success rates by 20% compared to conventional methods. Extensive evaluation on General-QA and IoT-Native datasets demonstrates the robustness of FHCR against adversarial threats, with FHCR maintaining global accuracy not lower than baseline levels while reducing communication costs by 14%.

Comparative analysis of ChatGPT-4o mini, ChatGPT-4o and Gemini Advanced in the treatment of postmenopausal osteoporosis

BackgroundOsteoporosis is a sex-specific disease. Postmenopausal osteoporosis (PMOP) has been the focus of public health research worldwide. The purpose of this study is to evaluate the quality and readability of artificial intelligence large-scale language models (AI-LLMs): ChatGPT-4o mini, ChatGPT-4o and Gemini Advanced for responses generated in response to questions related to PMOP.MethodsWe collected 48 PMOP frequently asked questions (FAQs) through offline counseling and online medical community forums. We also prepared 24 specific questions about PMOP based on the Management of Postmenopausal Osteoporosis: 2022 ACOG Clinical Practice Guideline No. 2 (2022 ACOG-PMOP Guideline). In this project, the FAQs were imported into the AI-LLMs (ChatGPT-4o mini, ChatGPT-4o, Gemini Advanced) and randomly assigned to four professional orthopedic surgeons, who independently rated the satisfaction of each response via a 5-point Likert scale. Furthermore, a Flesch Reading Ease (FRE) score was calculated for each of the LLMs’ responses to assess the readability of the text generated by each LLM.ResultsWhen it comes to addressing questions related to PMOP and the 2022 ACOG-PMOP guidelines, ChatGPT-4o and Gemini Advanced provide more concise answers than ChatGPT-4o mini. In terms of the overall FAQs of PMOP, ChatGPT-4o has a significantly higher accuracy rate than ChatGPT-4o mini and Gemini Advanced. When answering questions related to the 2022 ACOG-PMOP guidelines, ChatGPT-4o mini vs. ChatGPT-4o have significantly higher response accuracy than Gemini Advanced. ChatGPT-4o mini, ChatGPT-4o, and Gemini Advanced all have good levels of self-correction.ConclusionsOur research shows that Gemini Advanced and ChatGPT-4o provide more concise and intuitive answers. ChatGPT-4o responds better in answering frequently asked questions related to PMOP. When answering questions related to the 2022 ACOG-PMOP guidelines, ChatGPT-4o mini and ChatGPT-4o responded significantly better than Gemini Advanced. ChatGPT-4o mini, ChatGPT-4o, and Gemini Advanced have demonstrated a strong ability to self-correct.Clinical trial numberNot applicable.

Six Theses on Two Contemporary Literary Histories

The article mainly examines two three-volume literary-historical works – “Bulgarian Literature from the Liberation to the First World War” by Milena Kirova and “Literacy and Literature in Bulgaria IX – XXI Century” by Valeri Stefanov et. al., including two more volumes “Bulgarian Literature in the XXI Century (2000 – 2022)” again by M. Kirova – as large-scale and presentable models for making national literary histories in this globalist century. The main goal of the study is to derive six theses – on overcoming easy and unilinear historicism, on the different heterogeneity of methodological paradigms, on the specificity in structure and languages, on the changing relation to the canon and the periphery, on new interpretations or divergent readings, on the “illustrative” plan as a supplementary text for reading, which clearly show the correlations between the tactical pushback from traditions and the partial gravitation towards the new taste of the times.

End-to-End Multi-Modal Speaker Change Detection with Pre-Trained Models

In this work, we propose a multi-modal speaker change detection (SCD) approach with focal loss, which integrates both audio and text features to enhance detection performance. The proposed approach utilizes pre-trained large-scale models for feature extraction and incorporates a self-attention mechanism to optimize useful features related to speaker change. The extracted features are fused and processed through a fully connected classification network, with layer normalization and dropout for stability and generalization. To address class imbalance, we apply focal loss, which reduces errors for the difficult samples, leading to better balanced performance. Extensive experiments on a multi-talker meeting dataset demonstrate that the proposed multi-modal approach consistently outperforms single-modal models, proving the complementary nature of audio and text for SCD. Fine-tuning pre-trained models (Wav2Vec2 and Bert) for audio and text significantly boosts accuracy, achieving a 21% improvement over frozen models. The self-attention mechanism further improves performance by 2%, highlighting its ability to capture speaker transition cues effectively. Additionally, focal loss enhances the model’s performance, making it more robust to imbalanced data.

Stability Analysis of a Water Diversion Tunnel in Fractured Surrounding Rock via 3D-Printed Large-Scale Model Tests

Stability Analysis of a Water Diversion Tunnel in Fractured Surrounding Rock via 3D-Printed Large-Scale Model Tests

STLC-KG:A Social Text Steganalysis Method Combining Large-Scale Language Models and Common-Sense Knowledge Graphs

Language steganography in social networks primarily focuses on embedding secret information into social media text efficiently to achieve covert communication. The misuse of such techniques could pose significant potential threats to public cyberspace, such as the spread of malicious code, commands, or viruses. Existing social text steganalysis techniques mainly focus on the analysis of individual social media texts. However, the information content in a single text is very limited, leading to poor detection performance in practical applications. To address this challenge, this paper proposes a social text steganalysis method that combines large-scale language models with common-sense knowledge graphs (STLC-KG). This method first uses knowledge graphs to expand the knowledge contained in the text under investigation, enriching its linguistic expression, and then utilizes large-scale language models to extract the linguistic features of the social text. The results of tests conducted on three mainstream social media platforms demonstrate that the proposed method significantly improves the performance of social text steganalysis.

Feasibility Study for Multimodal Image-Based Assessment of Patient-Specific Intracranial Arteriovenous Malformation Hemodynamics.

Background/Objectives: Intracranial arteriovenous malformations (AVMs) exhibit a complex vasculature characterized by a locally occurring tangled nidus connecting the arterial and venous system bypassing the capillary network. Clinically available imaging modalities may not give sufficient spatial or temporal resolution. Adequate 3D models of large vascular areas and a detailed blood flow analysis of the nidus including the surrounding vessels are not available yet. Methods: Three representative AVM cases containing multimodal image data (3D rotational angiography, magnetic resonance angiography, magnetic resonance venography, and phase-contrast quantitative magnetic resonance imaging) are investigated. Image segmentation results in partial 3D models of the different vascular segments, which are merged into large-scale neurovascular models. Subsequently, image-based blood flow simulations are conducted based on the segmented models using patient-specific flow measurements as boundary conditions. Results: The segmentation results provide comprehensive 3D models of the overall arteriovenous morphology including realistic nidus vessels. The qualitative results of the hemodynamic simulations show realistic flow behavior in the complex vasculature. Feeding arteries exhibit increased wall shear stress (WSS) and higher flow velocities in two cases compared to contralateral vessels. In addition, feeding arteries are exposed to higher overall WSS with increased value variation between individual vessels (20.1 Pa ± 17.3 Pa) compared to the draining veins having a 62% lower WSS (8.9 Pa ± 5.9 Pa). Blood flow distribution is dragged towards the dominating circulation side feeding the nidus for all the cases quantified by the volume flow direction changes in the posterior communicating arteries. Conclusions: This multimodal study demonstrates the feasibility of the presented workflow to acquire detailed blood flow predictions in large-scale AVM models based on complex image data. The hemodynamic models serve as a base for endovascular treatment modeling influencing flow patterns in distally located vasculatures.

DiffuseHigh: Training-Free Progressive High-Resolution Image Synthesis Through Structure Guidance

Large-scale generative models, such as text-to-image diffusion models, have garnered widespread attention across diverse domains due to their creative and high-fidelity image generation. Nonetheless, existing large-scale diffusion models are confined to generating images of up to 1K resolution, which is far from meeting the demands of contemporary commercial applications. Directly sampling higher-resolution images often yields results marred by artifacts such as object repetition and distorted shapes. Addressing the aforementioned issues typically necessitates training or fine-tuning models on higher-resolution datasets. However, this poses a formidable challenge due to the difficulty in collecting large-scale high-resolution images and substantial computational resources. While several preceding works have proposed alternatives to bypass the cumbersome training process, they often fail to produce convincing results. In this work, we probe the generative ability of diffusion models at higher resolution beyond their original capability and propose a novel progressive approach that fully utilizes generated low-resolution images to guide the generation of higher-resolution images. Additionally, we integrate an image sharpening operation into our pipeline, further enhancing image quality. Our method obviates the need for additional training or fine-tuning which significantly lowers the burden of computational costs. Extensive experiments and results validate the efficiency and efficacy of our method.

Apollo-Forecast: Overcoming Aliasing and Inference Speed Challenges in Language Models for Time Series Forecasting

Encoding time series into tokens and using language models for processing has been shown to substantially augment the models' ability to generalize to unseen tasks. However, existing language models for time series forecasting encounter several obstacles, including aliasing distortion and prolonged inference times, primarily due to the limitations of quantization processes and the computational demands of large models. This paper introduces Apollo-Forecast, a novel framework that tackles these challenges with two key innovations: the Anti-Aliasing Quantization Module (AAQM) and the Race Decoding (RD) technique. AAQM adeptly encodes sequences into tokens while mitigating high-frequency noise in the original signals, thus enhancing both signal fidelity and overall quantization efficiency. RD employs a draft model to enable parallel processing and results integration, which markedly accelerates the inference speed for long-term predictions, particularly in large-scale models. Extensive experiments on various real-world datasets show that Apollo-Forecast outperforms state-of-the-art methods by 35.41% and 18.99% in WQL and MASE metrics, respectively, in zero-shot scenarios. Furthermore, our method achieves an acceleration of 1.9X-2.7X in inference speed over the baseline methods.

On the Consideration of AI Openness: Can Good Intent Be Abused?

Open source is a driving force behind scientific advancement. However, this openness is also a double-edged sword, with the inherent risk that innovative technologies can be misused for purposes harmful to society. What is the likelihood that an open source AI model or dataset will be used to commit a real-world crime, and if a criminal does exploit it, will the people behind the technology be able to escape legal liability? To address these questions, we explore a legal domain where individual choices can have a significant impact on society. Specifically, we build the EVE-v1 dataset that comprises 200 question-answer pairs related to criminal offenses based on 200 Korean precedents first to explore the possibility of malicious models emerging. We further developed EVE-v2 using 600 fraud-related precedents to confirm the existence of malicious models that can provide harmful advice on a wide range of criminal topics to test the domain generalization ability. Remarkably, widely used open-source large-scale language models (LLMs) provide unethical and detailed information about criminal activities when fine-tuned with \oursall. We also take an in-depth look at the legal issues that malicious language models and their builders could realistically face. Our findings highlight the paradoxical dilemma that open source accelerates scientific progress, but requires great care to minimize the potential for misuse.

Compression-Aware Computing for Scalable and Sustainable AI

This talk explores the challenge of customizing large-scale AI models, particularly generative AI, on cost-effective devices with limited memory and energy resources. Modern AI models demand substantial computational power, often relying on specialized hardware such as GPUs. To address this, the talk introduces compression-aware computing, a framework enabling AI models to recognize and adapt to their compressed states while preserving performance. Compression-aware computing integrates compression techniques like sparsification, quantization, and low-rank decomposition to enhance the efficiency and accuracy of AI models, broadening these models' accessibility across diverse devices. Additionally, this talk highlights one rationale of scalable and sustainable AI in advancing Alzheimer’s research by facilitating the analysis of large single-cell transcriptomics datasets for gene-gene interaction discovery.

Artificial Intelligence in Atrial Fibrillation: From Early Detection to Precision Therapy.

Atrial fibrillation (AF) is the most prevalent cardiac arrhythmia, associated with significant morbidity, mortality, and healthcare burden. Despite advances in AF management, challenges persist in early detection, risk stratification, and treatment optimization, necessitating innovative solutions. Artificial intelligence (AI) has emerged as a transformative tool in AF care, leveraging machine learning and deep learning algorithms to enhance diagnostic accuracy, improve risk prediction, and guide therapeutic interventions. AI-powered electrocardiographic screening has demonstrated the ability to detect asymptomatic AF, while wearable photoplethysmography-based technologies have expanded real-time rhythm monitoring beyond clinical settings. AI-driven predictive models integrate electronic health records and multimodal physiological data to refine AF risk stratification, stroke prediction, and anticoagulation decision making. In the realm of treatment, AI is revolutionizing individualized therapy and optimizing anticoagulation management and catheter ablation strategies. Notably, AI-enhanced electroanatomic mapping and real-time procedural guidance hold promise for improving ablation success rates and reducing AF recurrence. Despite these advancements, the clinical integration of AI in AF management remains an evolving field. Future research should focus on large-scale validation, model interpretability, and regulatory frameworks to ensure widespread adoption. This review explores the current and emerging applications of AI in AF, highlighting its potential to enhance precision medicine and patient outcomes.

Adaptive Prompt-Based Semantic Embedding with Inspire Potential of Implicit Knowledge for Cross-Modal Retrieval

In the era of big data, cross-modal retrieval is increasingly important in research and application. Given the latent complexity and non-intuitive nature of cross-modal relationships, leveraging external knowledge such as large models has become a popular approach to facilitate modality alignment. Existing methods typically address these challenges by fine-tuning model encoders or using a fixed number of prompts. However, these approaches struggle with the significant information asymmetry between image-text pairs and the high distribution diversity of image data. These limitations not only introduce noise during training but also reduce the accuracy and generalization capabilities in cross-modal retrieval tasks. To address the above issues, this paper proposes Adaptive Prompt-Based Semantic Embedding with Inspired Potential of Implicit Knowledge (APSE-IPIK). On one hand, we propose an inspired potential strategy to extract fine-grained and multi-perspective text descriptions from large-scale pre-trained multimodal models, which can be seen as implicit knowledge injection. These descriptions are integrated into the visual-semantic embedding through cross-modal semantic alignment with images, balancing the information asymmetry between modalities and reducing the embedding of inaccurate mapping relationships. On the other hand, we construct an instance-level query-based prompt pool strategy to adaptively extract the most relevant prompts, addressing alignment biases caused by intra-modal (especially image) data diversity and improving alignment accuracy. Extensive experiments are conducted on two widely used datasets, Flickr30k and MSCOCO, which show the effectiveness of the proposed method.

GenesisTex2: Stable, Consistent and High-Quality Text-to-Texture Generation

Large-scale text-guided image diffusion models have demonstrated remarkable results in text-to-image (T2I) generation. However, applying these models to synthesize textures for 3D geometries remains challenging due to the domain gap between 2D images and textures on a 3D surface. Early works that used a projecting-inpainting approach managed to preserve generation diversity, but often resulted in noticeable artifacts and style inconsistencies. While recent methods have attempted to address these inconsistencies, they often introduce other issues, such as blurring, over-saturation, or over-smoothing. To overcome these challenges, we propose a novel text-to-texture synthesis framework that takes advantage of pre-trained diffusion models. We introduce a local attention reweighing mechanism in the self-attention layers to guide the model in focusing on spatial-correlated patches across different views, thereby enhancing local details while preserving cross-view consistency. Additionally, we propose a novel latent space merge pipeline, which further ensures consistency across different viewpoints without sacrificing too much diversity. Our method significantly outperforms existing state-of-the-art techniques in terms of texture consistency and visual quality, while delivering results much faster than distillation-based methods. Importantly, our framework does not require additional training or fine-tuning, making it highly adaptable to a wide range of models available on public platforms.

FashionTailor: Controllable Clothing Editing for Human Images with Appearance Preserving

The garment structure serves as a crucial medium for expressing the designer's creative vision and showcasing the distinctive character of clothing items. Effective editing of garment structure in fashion images allows for an advanced preview of the design, accelerating the process of garment customization to meet individualized requirements. Although large-scale diffusion models have demonstrated impressive image generation and editing capabilities, no efforts have been made to exploit their potential in part-level editing of images. Unlike previous research, we define a clothing structure editing (CSE) task aimed at accurately editing the local structure of human-centered clothing images through simple instruction-based prompts while maintaining the consistency of clothing appearance. Specifically, this paper develops a new controllable triple-flow framework for structure editing named FashionTailor. An additional network called ClothingNet is proposed to extract the clothing details to address the rigid constraints of the original garment structure. Then, we propose a semantic-refined module to extract the semantic understanding of the source image and adaptively focus on the part to be edited. We also design a cross-blend attention mechanism to integrate fine-grained clothing features to guarantee precise alignment between appearance and target structure features. In addition, a garment structure dataset called StructureFashion has been collated, wherein each item of clothing is represented by multiple photos with diverse structure characteristics, containing over six million pairs. Finally, our method supports editing the structure of multiple parts on a garment simultaneously. Extensive experiments validate the effectiveness of our method for editing part-level human images in StructureFashion dataset and real-scenarios.

MUSE: Mamba Is Efficient Multi-scale Learner for Text-video Retrieval

Text-Video Retrieval (TVR) aims to align and associate relevant video content with corresponding natural language queries. Most existing TVR methods are based on large-scale pre-trained vision-language models (e.g., CLIP). However, due to CLIP's inherent plain structure, few TVR methods explore the multi-scale representations which offer richer contextual information for a more thorough understanding. To this end, we propose MUSE, a multi-scale mamba with linear computational complexity for efficient cross-resolution modeling. Specifically, the multi-scale representations are generated by applying a feature pyramid on the last single-scale feature map. Then, we employ the Mamba structure as an efficient multi-scale learner to jointly learn scale-wise representations. Furthermore, we conduct comprehensive studies to investigate different model structures and designs. Extensive results on three popular benchmarks have validated the superiority of MUSE.

Cobra: Extending Mamba to Multi-Modal Large Language Model for Efficient Inference

In recent years, applying multi-modal large language models (MLLMs) in various fields has achieved remarkable success. However, as the foundation model for many downstream tasks, MLLMs comprise the well-known Transformer network, which has a less efficient quadratic computation complexity. In this study, we introduce Cobra, a multi-modal large-scale language model built upon a state-space model, which has demonstrated significant potential in efficiently handling long sequences with fast inference and linear scalability concerning sequence length. Specifically, Cobra involves replacing Transformer-based backbone models (e.g., LLaMA or Phi) with pre-trained Mamba language models. We then empirically explore effective strategies for aligning visual and textual modalities and integrating various pre-trained Mamba model variants with visual encoders. Experiments across various multi-modal benchmarks demonstrate that: (i) Cobra performs 3× ∼ 4× faster than the most computationally efficient state-of-the-art methods, e.g., LLaVA-Phi and MobileVLM v2. Additionally, its performance is significantly enhanced thanks to the implementation of linear sequential modeling. (ii) Cobra fine-tunes a small parameter (∼48% of model parameters), leading to a significant improvement in overall performance compared to LLaVA.

MambaPro: Multi-Modal Object Re-identification with Mamba Aggregation and Synergistic Prompt

Multi-modal object Re-IDentification (ReID) aims to retrieve specific objects by utilizing complementary image information from different modalities. Recently, large-scale pre-trained models like CLIP have demonstrated impressive performance in traditional single-modal ReID tasks. However, they remain unexplored for multi-modal object ReID. Furthermore, current multi-modal aggregation methods have obvious limitations in dealing with long sequences from different modalities. To address above issues, we introduce a novel framework called MambaPro for multi-modal object ReID. To be specific, we first employ a Parallel Feed-Forward Adapter (PFA) for adapting CLIP to multi-modal object ReID. Then, we propose the Synergistic Residual Prompt (SRP) to guide the joint learning of multi-modal features. Finally, leveraging Mamba's superior scalability for long sequences, we introduce Mamba Aggregation (MA) to efficiently model interactions between different modalities. As a result, MambaPro could extract more robust features with lower complexity. Extensive experiments on three multi-modal object ReID benchmarks (i.e., RGBNT201, RGBNT100 and MSVR310) validate the effectiveness of our proposed methods.