Researchers at Nanjing University have developed a new training framework and benchmark that dramatically enhance how AI communicates and coordinates with humans in unpredictable, real-world scenarios, paving the way for safer, more efficient collaboration across various industries.
Researchers developed FOCUS, an offline RL framework that integrates causal structured world models to distinguish true causal relationships from spurious correlations in historical data. Theoretically, this approach tightens generalization error bounds, while experiments show it outperforms existing methods on benchmark tasks.
Researchers have developed a drone system that uses blockchain sharding to significantly enhance the speed, security, and real-time coordination of search and rescue missions.
Researchers have developed an innovative AI-driven system that automates and optimizes penetration testing, significantly reducing testing steps while expanding vulnerability assessments for large-scale networks.
Researchers at Shanghai Jiao Tong University have developed CHASER, a blockchain-based incentive system that transforms mobile crowdsensing by leveraging automated smart contracts and advanced encryption to deliver fair compensation, robust data security, and high efficiency, dramatically boosting user participation and reliability.
Researchers at Wuhan University have introduced Fair Adversarial Training (FairAT), a technique that enhances AI security and fairness by identifying and strengthening vulnerable data points, thereby reducing cyber-attack risks and mitigating ethical biases.
Researchers at Soochow University have unveiled DPEC, an advanced AI model that uses dual-view prompts and element correlation to significantly enhance spatial relation extraction from text, setting a new benchmark for technologies in autonomous driving, digital mapping, and smart personal assistants.
Researchers from Jinan University, Huawei, and ByteDance have developed LTAA-FGAC, an innovative authentication system that balances user privacy with public traceability and fine-grained access control to enhance digital security and accountability.
Researchers at Harbin Institute of Technology and Singapore Management University have developed LR-GCN, a method that leverages reinforcement learning, logical reasoning, and advanced graph neural networks to boost AI accuracy by up to 17% on sparse datasets, effectively addressing the challenge of Knowledge Graph Completion in critical industries.
Dynamic-EC, developed by researchers at Shanghai Jiao Tong University, introduces a smarter approach to blockchain storage by using real-time risk assessment to drastically lower costs and improve performance.
A research collaboration between Shanghai Jiao Tong University, Shanghai Qi Zhi Institution, and Huawei Technologies has introduced 鈥淏AFT鈥�, a cutting-edge auto-save system for AI training that minimizes downtime and optimizes efficiency. Designed to leverage idle moments in training workflows, BAFT significantly enhances fault tolerance while reducing computational overhead, setting a new industry benchmark for reliable AI model development.
Researchers from the National University of Defense Technology (China) have developed an innovative processor that enables seamless execution of legacy software on next-generation hardware without costly redevelopment. This breakthrough technology integrates hardware-based translation, allowing software originally designed for ARM-based chips to run efficiently on the open-source RISC-V architecture while maintaining up to 65% of native performance. By eliminating software compatibility barriers, this solution offers a cost-effective and power-efficient path for industries relying on embedded systems, including smart devices, medical sensors, and industrial automation.
A new study reveals that 94% of business spreadsheets contain errors, leading to significant risks.
Researchers from Islamic Azad University have developed innovative designs for quantum circuits that reduce costs by over 25% and significantly enhance error detection. These advancements aim to improve the efficiency and reliability of quantum computing.
MARISMA is a risk analysis framework designed to improve cybersecurity for businesses by offering adaptive, real-time protection for digital assets against evolving threats.
A novel encrypt-then-index strategy improves secure data queries on encrypted cloud databases, offering enhanced efficiency and privacy-preserving machine learning.
Graph-Segmenter improves image segmentation with graph transformers and boundary-aware attention, boosting road safety and medical diagnostics.
Researchers have developed the Interactive Relation Embedding Network (IRE-Net), an AI system that can detect contactless human interactions in crowded settings, enhancing tools for public safety and social behavior analysis.
Researchers from Hubei Minzu University and Wuhan University have further advanced CRD-CGAN, an AI model that generates detailed and varied images from text, offering enhanced applications in sectors like digital marketing and virtual reality.
The newly developed Joint Matrix Decomposition and Factorization (JMDF) framework improves the accuracy and robustness of moving target detection in video streams through the use of fuzzy logic and adaptive constraints.
RSS
Medicine keyboard_arrow_right
Sciencekeyboard_arrow_right
Life keyboard_arrow_right
Business keyboard_arrow_right
Journal 蜜桃传媒 keyboard_arrow_right
By Location keyboard_arrow_right
Meetings, Grants, and Events keyboard_arrow_right