Researchers introduced LEXIC, a lightweight method that improves gaze-only reading comprehension models by incorporating word difficulty metrics without relying on heavy language models.
Researchers introduced GAP-GDRNet, a geometry-aware monocular 6D pose estimation framework designed to improve spacecraft tracking under challenging lighting and structural conditions.
CommuniWave is a machine learning model developed to measure and analyze informal resident behaviors to assist in resilient urban planning.
A new reconstruction method has been developed to improve the accuracy of oxygen saturation estimation from low-quality wearable sensor signals.
Accepted papers at the International Conference on Machine Learning demonstrate that open-source frontier models and infrastructure have become central to modern AI research.
The authors propose a method for building knowledge graphs and detailed profiles of academic conferences to help researchers navigate scientific events and publications.
This paper discusses methods for constructing accurate digital profiles of scientific and technological resources to improve knowledge services in the cloud computing era.
This paper presents mathematical formulations for three foundational problems in classical mechanics using a complex-time representation.
A new deep learning network improves automatic modulation classification across different communication environments by leveraging signal priors and domain adaptation.
A new study details VectorizationLLM, a specialized model built on Google's open weights to help engineering students learn computational analysis in MATLAB.
Researchers have designed a deep learning technique to approximate the stationary distribution and performance metrics of high-dimensional reflected Brownian motion systems.
This paper presents a retrieval system designed to map and search the scientific achievements and expertise of academic researchers.
The paper introduces a deep learning-based noise removal method called RAB-U-Net to improve the accuracy of engine sound analysis during manufacturing quality control.
Researchers developed a fuzzy inference system to dynamically scale validator nodes in private blockchain networks based on real-time workload.
The paper explores how an autogenerative theory of language can address theoretical gaps in integrationist linguistics regarding how signs sustain communication.
The successful demonstration of reusable rocket capabilities by China's state-owned Long March 10B raises existential questions for the country's private commercial rocket startups.
Hugging Face has published the fourth installment of its PRX series, detailing the data strategy used for training its models.
Researchers at ByteDance have identified a new scaling law showing that autonomous AI agents can double their learning speed quarterly through real-world task interactions.
Google DeepMind has partnered with entertainment company A24 to explore the collaborative use of artificial intelligence in filmmaking and creative processes.
Alibaba's Damo Academy has introduced an AI agent called Elements Claw, which successfully identified four previously unknown superconducting compounds that were later verified in a laboratory.
MIT physics professor Jesse Thaler, who leads the NSF AI Institute for Artificial Intelligence and Fundamental Interactions, has been appointed as the new director of the Laboratory for Nuclear Science.
A UN report indicates that China leads the world in power technology innovation and ranks near the top in overall power development.
Shanghai has established a dedicated quantum computing incubation zone to accelerate the development of quantum technologies and foster collaboration among twenty-six initial firms.
Chinese scientists have developed advanced thermal insulation materials capable of withstanding temperatures over 1000 degrees Celsius for reusable rocket applications.