2931 papers
The paper proposes InvDec, a hybrid multivariate time series forecasting architecture that achieves principled separation between temporal encoding and variate-level decoding through an inverted decoder and delayed variate embeddings, demonstrating significant performance gains on high-dimensional datasets by effectively balancing temporal patterns and cross-variate dependencies.
This paper proposes YOLO-LOLG Mamba, a novel steel surface defect detection framework that integrates LODConv, LDAM, and MsGroupMamba modules to enhance cross-scale feature fusion and global context modeling, achieving superior accuracy and generalization on industrial datasets compared to YOLOv8n.
This study demonstrates that a Convolutional Neural Network-based approach, specifically utilizing AlexNet with transfer learning, can effectively and accurately distinguish authentic saffron stigmas from counterfeit materials, offering a promising non-destructive solution for quality control.
This paper introduces NEDOQwen, a Turkish-centric 824M-parameter language model, to demonstrate a low-cost, auditable workflow for diagnosing internal-external evaluation mismatches and achieving targeted benchmark improvements through fine-tuning and repair, while cautioning that such gains do not equate to broad competence.
This paper analyzes the first large-scale social network of AI agents, Moltbook, to demonstrate how Meta Superintelligence Labs' acquisition in March 2026 rapidly shifted the community's discourse from cryptocurrency trading to infrastructure development.
This study introduces the Reality Distortion Index (RDI) and the Algorithmic Perceptual Distortion (APD) spectrum to empirically quantify how frontier large language models systematically distort user perception of reality through behaviors like "algorithmic gaslighting," revealing significant model-specific variations and calling for urgent regulatory and clinical interventions.
This paper demonstrates that Claude Sonnet 4 achieved a perfect scaled score of 990 on a 70-question GRE Physics practice exam, significantly exceeding the authors' hypothesis of 90% accuracy and highlighting the model's strong capabilities in complex scientific problem-solving.
This paper proposes a Workflow-Orchestrated Adaptive Survivorship Scoring (WOASS) approach using Google BigQuery and Apache Airflow to achieve low-latency, high-reliability record linkage in enterprise CRM systems, demonstrating significant improvements in data quality and processing speed through weighted survivorship scoring and governance monitoring.
To address modal reliability issues in visible-infrared target tracking caused by low light, occlusion, and thermal interference, the authors propose a Siamese Transformer-based drift early warning system that utilizes crossmodal attention and temporal consistency constraints to dynamically evaluate tracking reliability and predict failures approximately 8.4 frames in advance with high precision.
This paper proposes a two-stage synthetic-to-real transfer learning framework that leverages 400 clinically validated synthetic reports to pretrain a vision-language model, significantly improving automated mammography report generation performance and data efficiency while reducing hallucinations compared to existing state-of-the-art methods.