Automatic X-ray prohibited items detection is vital for security inspection and has been widely studied. Traditional methods rely on visual modality, often struggling with complex threats. While recent studies incorporate language to guide single-view images, human inspectors typically use dual-view images in practice. This raises the question: can the second view provide constraints similar to a language modality? In this work, we introduce DualXrayBench, the first comprehensive benchmark for X-ray inspection that includes multiple views and modalities. It supports eight tasks designed to test cross-view reasoning. In DualXrayBench, we introduce a caption corpus consisting of 45,613 dual-view image pairs across 12 categories with corresponding captions. Building upon these data, we propose the Geometric (cross-view)-Semantic (cross-modality) Reasoner (GSR), a multimodal model that jointly learns correspondences between cross-view geometry and cross-modal semantics, treating the second-view images as a "language-like modality". To enable this, we construct the GSXray dataset, with structured Chain-of-Thought sequences: , , . Comprehensive evaluations on DualXrayBench demonstrate that GSR achieves significant improvements across all X-ray tasks, offering a new perspective for real-world X-ray inspection.

The rise of Generative AI Search is fundamentally transforming how users and intelligent systems interact with the Internet. LLMs increasingly act as intermediaries between humans and web information. Yet the web remains optimized for human browsing rather than AI-driven semantic retrieval, resulting in wasted network bandwidth, lower information quality, and unnecessary complexity for developers. We introduce the concept of an AI-Native Internet, a web architecture in which servers expose semantically relevant information chunks rather than full documents, supported by a Web-native semantic resolver that allows AI applications to discover relevant information sources before retrieving fine-grained chunks. Through motivational experiments, we quantify the inefficiencies of current HTML-based retrieval, and outline architectural directions and open challenges for evolving today's document-centric web into an AI-oriented substrate that better supports semantic access to web content.

The ability of Large Language Models (LLMs) to generate structured outputs that follow arbitrary schemas is crucial to a wide range of downstream tasks that require diverse structured representations of results such as information extraction, table generation, and function calling. While modern LLMs excel in generating unstructured responses in natural language, whether this advancement translates to a strong performance on text-to-structure tasks remains unclear. To bridge this gap, we first introduce OmniStruct, a comprehensive benchmark for assessing LLMs' capabilities on diverse text-to-structure tasks such as information extraction, table generation, and function calling. We build OmniStruct by identifying existing datasets across a wide range of tasks that are suitable for a structured answer format, and adapting them under a unified text-to-structure problem setting. To facilitate the development of efficient text-to-structure models, we collect high-quality training data via synthetic task generation. Without using any supervised data for OmniStruct tasks, our experiments demonstrate the possibility of fine-tuning much smaller models on synthetic data into universal structured generation models that can rival the performance of GPT-4o.

Brain tumor detection from MRI scans plays a crucial role in early diagnosis and treatment planning. Deep convolutional neural networks (CNNs) have demonstrated strong performance in medical imaging tasks, particularly when pretrained on large datasets. However, it remains unclear which type of pretrained model performs better when only a small dataset is available: those trained on domain-specific medical data or those pretrained on large general datasets. In this study, we systematically evaluate three pretrained CNN architectures for brain tumor classification: RadImageNet DenseNet121 with medical-domain pretraining, EfficientNetV2S, and ConvNeXt-Tiny, which are modern general-purpose CNNs. All models were trained and fine-tuned under identical conditions using a limited-size brain MRI dataset to ensure a fair comparison. Our results reveal that ConvNeXt-Tiny achieved the highest accuracy, followed by EfficientNetV2S, while RadImageNet DenseNet121, despite being pretrained on domain-specific medical data, exhibited poor generalization with lower accuracy and higher loss. These findings suggest that domain-specific pretraining may not generalize well under small-data conditions. In contrast, modern, deeper general-purpose CNNs pretrained on large-scale datasets can offer superior transfer learning performance in specialized medical imaging tasks.

This paper introduces the approach of "Gradient Masters" for BLP-2025 Task 1: "Bangla Multitask Hate Speech Identification Shared Task". We present an ensemble-based fine-tuning strategy for addressing subtasks 1A (hate-type classification) and 1B (target group classification) in YouTube comments. We propose a hybrid approach on a Bangla Language Model, which outperformed the baseline models and secured the 6th position in subtask 1A with a micro F1 score of 73.23% and the third position in subtask 1B with 73.28%. We conducted extensive experiments that evaluated the robustness of the model throughout the development and evaluation phases, including comparisons with other Language Model variants, to measure generalization in low-resource Bangla hate speech scenarios and data set coverage. In addition, we provide a detailed analysis of our findings, exploring misclassification patterns in the detection of hate speech.

Deep learning training relies on periodic checkpoints to recover from failures, but unsafe checkpoint installation can leave corrupted files on disk. This paper presents an experimental study of checkpoint installation protocols and integrity validation for AI training on macOS/APFS. We implement three write modes with increasing durability guarantees: unsafe (baseline, no fsync), atomic_nodirsync (file-level durability via fsync()), and atomic_dirsync (file + directory durability). We design a format-agnostic integrity guard using SHA-256 checksums with automatic rollback. Through controlled experiments including crash injection (430 unsafe-mode trials) and corruption injection (1,600 atomic-mode trials), we demonstrate that the integrity guard detects 99.8-100% of corruptions with zero false positives. Performance overhead is 56.5-108.4% for atomic_nodirsync and 84.2-570.6% for atomic_dirsync relative to the unsafe baseline. Our findings quantify the reliability-performance trade-offs and provide deployment guidance for production AI infrastructure.

Autonomous inspection in hazardous environments requires AI agents that can interpret high-level goals and execute precise control. A key capability for such agents is spatial grounding, for example when a drone must center a detected object in its camera view to enable reliable inspection. While large language models provide a natural interface for specifying goals, using them directly for visual control achieves only 58\% success in this task. We envision that equipping agents with a world model as a tool would allow them to roll out candidate actions and perform better in spatially grounded settings, but conventional world models are data and compute intensive. To address this, we propose a task-specific latent dynamics model that learns state-specific action-induced shifts in a shared latent space using only goal-state supervision. The model leverages global action embeddings and complementary training losses to stabilize learning. In experiments, our approach achieves 71\% success and generalizes to unseen images and instructions, highlighting the potential of compact, domain-specific latent dynamics models for spatial alignment in autonomous inspection.

Time series data plays a pivotal role in a wide variety of fields but faces challenges related to privacy concerns. Recently, synthesizing data via diffusion models is viewed as a promising solution. However, existing methods still struggle to capture long-range temporal dependencies and complex channel interrelations. In this research, we aim to utilize the sequence modeling capability of a State Space Model called Mamba to extend its applicability to time series data generation. We firstly analyze the core limitations in State Space Model, namely the lack of consideration for correlated temporal lag and channel permutation. Building upon the insight, we propose Lag Fusion Mamba and Permutation Scanning Mamba, which enhance the model's ability to discern significant patterns during the denoising process. Theoretical analysis reveals that both variants exhibit a unified matrix multiplication framework with the original Mamba, offering a deeper understanding of our method. Finally, we integrate two variants and introduce Diffusion Mamba for Time Series (DiM-TS), a high-quality time series generation model that better preserves the temporal periodicity and inter-channel correlations. Comprehensive experiments on public datasets demonstrate the superiority of DiM-TS in generating realistic time series while preserving diverse properties of data.

Styled handwriting generation aims to synthesize handwritten text that looks both realistic and aligned with a specific writer's style. While recent approaches involving GAN, transformer and diffusion-based models have made progress, they often struggle to capture the full spectrum of writer-specific attributes, particularly global stylistic patterns that span long-range spatial dependencies. As a result, capturing subtle writer-specific traits such as consistent slant, curvature or stroke pressure, while keeping the generated text accurate is still an open problem. In this work, we present a unified framework designed to address these limitations. We introduce a Vision Transformer-based style encoder that learns global stylistic patterns from multiple reference images, allowing the model to better represent long-range structural characteristics of handwriting. We then integrate these style cues with the target text using a cross-attention mechanism, enabling the system to produce handwritten images that more faithfully reflect the intended style. To make the process more interpretable, we utilize Salient Stroke Attention Analysis (SSAA), which reveals the stroke-level features the model focuses on during style transfer. Together, these components lead to handwriting synthesis that is not only more stylistically coherent, but also easier to understand and analyze.

The exponential growth of scientific knowledge has created significant barriers to cross-disciplinary knowledge discovery, synthesis and research collaboration. In response to this challenge, we present BioSage, a novel compound AI architecture that integrates LLMs with RAG, orchestrated specialized agents and tools to enable discoveries across AI, data science, biomedical, and biosecurity domains. Our system features several specialized agents including the retrieval agent with query planning and response synthesis that enable knowledge retrieval across domains with citation-backed responses, cross-disciplinary translation agents that align specialized terminology and methodologies, and reasoning agents that synthesize domain-specific insights with transparency, traceability and usability. We demonstrate the effectiveness of our BioSage system through a rigorous evaluation on scientific benchmarks (LitQA2, GPQA, WMDP, HLE-Bio) and introduce a new cross-modal benchmark for biology and AI, showing that our BioSage agents outperform vanilla and RAG approaches by 13\%-21\% powered by Llama 3.1. 70B and GPT-4o models. We perform causal investigations into compound AI system behavior and report significant performance improvements by adding RAG and agents over the vanilla models. Unlike other systems, our solution is driven by user-centric design principles and orchestrates specialized user-agent interaction workflows supporting scientific activities including but not limited to summarization, research debate and brainstorming. Our ongoing work focuses on multimodal retrieval and reasoning over charts, tables, and structured scientific data, along with developing comprehensive multimodal benchmarks for cross-disciplinary discovery. Our compound AI solution demonstrates significant potential for accelerating scientific advancement by reducing barriers between traditionally siloed domains.