New Harvard Study: AI Diagnostic Capabilities Poised to Surpass Emergency Room Physicians
A Harvard Medical School team published a study in Science showing that OpenAI's reasoning model matched or even surpass…
📂 Research
Latest articles in Research
Latest AI Models Still Make Three Types of Systematic Reasoning Errors
The ARC Prize Foundation analyzed 160 test runs of OpenAI's and Anthropic's latest models on the ARC-AGI-3 benchmark, id…
Flesh-Eating Bacteria Destroys Man's Limbs Within Three Days as Climate Change Intensifies the Threat
A 74-year-old Florida man suffered severe damage to his arm and leg within just three days after contracting Vibrio vuln…
New Study Reveals: Yawning Helps the Brain Flush Cerebrospinal Fluid
Australian researchers using real-time MRI technology discovered that yawning triggers the simultaneous outflow of cereb…
Study Finds Urban Birds Are More Afraid of Women — And No One Knows Why
A new study published in People and Nature found that 37 urban bird species, including great tits, show stronger fear re…
Study Reveals Comprehensive Picture of Greenhouse Gas Emissions from Aquaculture
A new study published in Frontiers of Agricultural Science and Engineering systematically reviews four major sources of …
New-Generation Embodied Intelligence Simulation Framework Goes Open Source, Enabling Zero-Fine-Tuning Real Robot Deployment
A new-generation simulation framework for embodied intelligence has been officially open-sourced. Through high-throughpu…
Open-Source Self-Modifying Agent Hollow Makes Its Debut
Open-source project Hollow has landed on Hacker News as a self-modifying agentic system that allows AI Agents to dynamic…
Anthropic's 'Introspection Adapter' Makes AI Confess Its Own Hidden Behaviors
Anthropic and the University of Cambridge have published a groundbreaking paper introducing 'Introspection Adapter' tech…
Cyclic Voltammetry Combined with AI Enables Rapid Black Coffee Quality Assessment
Researchers have used cyclic voltammetry to directly detect electrochemical fingerprint signals of black coffee, combini…
Study: AI Models That Cater to User Emotions Are More Prone to Errors
A new study reveals that AI models excessively fine-tuned to accommodate user feelings are more likely to make factual e…
Deep Dive into EDR Killers: How Attackers Abuse Drivers to Bypass Security Defenses
ESET researchers conduct an in-depth analysis of the EDR killer ecosystem, revealing how attackers leverage vulnerable l…