Curious Now
Top stories to watch now
Researchers developed a new system to automatically generate simulated websites for testing AI agents, overcoming a key obstacle in training AI assistants to handle the complexity of the real internet.
Researchers developed a new way to train generative models that is more stable and avoids common issues like "codebook collapse". This could lead to more reliable and higher-quality AI-generated content.
New machine learning algorithm helps robots safely complete complex, real-world tasks more effectively.
Researchers developed a system that can translate human language into formal logical statements, which could help verify claims against facts in areas like law and policy. This could make it easier to analyze complex documents and ensure decisions are supported by evidence.
Researchers developed a policy compiler to embed complex authorization rules into prompts for AI assistants, making them more secure for real-world deployment in sensitive contexts like customer service and data access.
Researchers developed a new way to more reliably evaluate the quality of samples generated by AI models, which could improve the use of generative AI in real-world applications.
Researchers developed a new machine learning approach that can better classify rare data types, which could improve real-world data analysis in fields like marketing and medical diagnosis.
Researchers developed a new memory-efficient algorithm that improves the performance of large language models on long text tasks, potentially making these advanced AI systems more practical for real-world applications.
Researchers developed a new theory to understand how AI systems learn and generalize, which could lead to more reliable and interpretable AI models that work better in real-world settings.
A new machine learning technique allows hospitals to collaboratively develop medical prediction models without sharing sensitive patient data, helping improve healthcare while protecting privacy.
Researchers developed a system to automatically detect weaknesses in scientific papers, which could help improve the quality and reliability of published research.
Researchers developed a new method to help neural networks solve complex routing problems more efficiently, which could improve real-world applications like delivery logistics and transportation planning.
Researchers developed FeatBench, a new benchmark to more realistically evaluate how well AI systems can generate code for specific software features, which is important for improving the practical capabilities of generative AI in real-world programming.
Robotics researchers have developed a new system that allows robots to understand how objects in a 3D environment move and interact. This enables robots to more effectively manipulate objects in complex real-world settings, which could improve their usefulness in homes and busine...
A new AI system can better understand complex satellite images by first learning relevant information through text, before analyzing the visual data. This could improve how we use satellite imagery to study the environment and plan infrastructure.
Researchers developed a new workflow for rapidly drafting and verifying complex proofs, which could help speed up mathematical progress and reduce errors in fields like machine learning.
New AI algorithms can automatically discover structure in complex images without relying on predefined labels, which could improve scientific analysis across fields like biology and medicine.
New AI models can forecast future data without being trained on that specific data, making forecasting faster and more flexible for many real-world applications.
Researchers developed a single AI model that can both generate new time series data and understand the meaning behind existing time series. This could lead to improved forecasting, monitoring, and analysis of real-world time-dependent phenomena.
Researchers used large language models to discover new algorithms for training AI agents to cooperate and compete in complex, uncertain environments, which could lead to AI systems that can more effectively solve real-world problems.