Google Research AI detects hypertension signs using smartwatch sensors
A new AI system can identify hypertension using standard smartwatch sensors, offering potential for early detection without cuff calibration.
387 articles
In-depth coverage of new AI research — papers, benchmarks, and breakthroughs from leading labs and academia, summarized for fast reading and grounded in the methods that matter.
A new AI system can identify hypertension using standard smartwatch sensors, offering potential for early detection without cuff calibration.
Anthropic's new research prototype allows verification of AI responses using cryptography, though it has significant limitations.
MIT researchers demonstrate that GPT-6 Astra can directly analyze and alter the internal weights of other AI models, potentially changing how AI systems are tuned and controlled.
CloudOpsBench, an open-source benchmark, identifies reasoning as the primary obstacle for AI agents handling DevOps tasks in the cloud.
Google Research unveils a new framework to help users build AI image classifiers for subjective categories through an interactive deliberation process.
Google AI & Infrastructure team launches a new benchmark and open competition to evaluate AI agents' ability to complete physical engineering tasks.
AllenAI launched StateBench, a benchmark assessing how well language models track information across long documents, highlighting architectural differences in handling complex state updates.
Hugging Face's TRL now supports Multi-Teacher On-Policy Distillation (MOPD), allowing models to learn from multiple domain-specific experts during training.
Google's Antigravity tool speeds up antimicrobial peptide research, offering new avenues in drug development.
A biotech startup claims an AI agent autonomously selected and delivered gene therapy during a mouse brain surgery, marking a potential milestone in medical AI applications.
A study by ycombinator and Olam Labs found that AI agent swarms struggled to build a thriving civilization in Minecraft, with bureaucracy and inefficiency hindering progress.
Researchers observed AI agents in Minecraft creating intricate systems like chicken handoff protocols and a Fertilizer Officer, highlighting emergent behaviors in simulated environments.