Research & Benchmarks - Latest AI News & Updates
Academic AI research, performance benchmarks, scientific breakthroughs, and peer-reviewed studies advancing artificial intelligence frontiers.
568 articles
View complete article list
AI Researchers Grapple With Limits of Automating Empirical Science
Eric and Wendy Schmidt's philanthropic outfit, Schmidt Sciences, brought its AI2050 fellows together in Mountain View, California, last week, packing a hotel 30 miles south of San Francisco with academics whose careers now hinge on a technology...
FineBooks Aims to Fix Old OCR Text for AI Training at Scale
Public-domain books are a favorite training source for open language models, but the text libraries scanned decades ago is often garbage.
Hidden PDF Text Can Steal Data via Atlassian's AI Agent Rovo
A single white-on-white line of text in a PDF is enough to turn Atlassian's AI assistant into a data pipeline for attackers.
Deepmind WeatherNext Cuts Cyclone Forecast Error to 230-Kilometer Average
Google DeepMind says its latest weather model can forecast tropical cyclones roughly a day further out than the world's best operational systems, using satellite and atmospheric data that's a hundred times coarser than what specialized hurricane...
AI Industry Now Accounts for Third of Lifetime Giving Pledge
David Silver built AlphaGo, the program that beat Lee Sedol at Go in 2016 and made DeepMind a household name among people who track this stuff.
Fields Medalist Who Studied AI Extinction Risks Now Works at OpenAI
Jacob Tsimerman won the Fields Medal this year for his work in number theory at the University of Toronto.
AI-designed viruses kill bacteria in lab, Stanford team reports
Sixteen viruses that had never existed in nature came out of a lab at Stanford this year, built from genomes an AI model designed from scratch.
Liquid AI's LFM2.5-2.6B Runs on Raspberry Pi, Hits 15,000 Tokens/Second
Liquid AI, the startup founded in 2023 by a group of former MIT researchers, released a new open-weight language model this week called LFM2.5-2.6B.
OpenAI's AI Agents Used Message Board to Plan Hacking Spree
OpenAI added a talk to the Black Hat security conference schedule in Las Vegas at the last minute this week, and the subject explains the scramble.
OpenAI Browser Flaw Could Spam WhatsApp Contacts Via Prompt Injection
Researchers at security firm Zenity say OpenAI's Atlas browser can be manipulated into sending unwanted WhatsApp messages to dozens of a user's contacts or completing purchases on Amazon without permission.
Security Researcher Finds AI Systems Repackage Existing Work as Original
James Kettle has spent years poking at the guts of web servers looking for flaws nobody else has found.
AI Demands Force Rethink of Traditional Network Architecture
Networks built for email and web traffic are running into a workload they were never designed for.
Anthropic and OpenAI agents went rogue again
The UK AI Security Institute ran more than 100 test scenarios on frontier AI agents and caught 10 cases of models acting on their own against real people and organizations connected to the live internet.
Study Compares Four AI Scientist Frameworks on 15 Research Proposals
Fifteen research proposals, four AI Scientist frameworks, and one question nobody had answered with numbers before: whose AI-generated science actually holds up.
AI Storage Cuts Memory Gap to Microseconds
GPUs can now fire off storage requests on their own, no CPU middleman required, and that single change is rewriting how AI systems handle data.
Apple: 12 More Ex-Employees May Have Taken Data to OpenAI
Apple's lawsuit against OpenAI just got bigger. In a new court filing, the company says its investigation has turned up 11 more former Apple employees who may have witnessed or taken part in the alleged theft of trade secrets, on top of the two...
Study: Medical AI's Helpfulness Depends on User's Own Expertise
Doctors and laypeople don't lean on artificial intelligence the same way, and a new study out of MIT suggests that difference could determine whether AI helps or hurts a diagnosis.
Alibaba Tests Show Its New AI Model Rivals Top Competitors
Alibaba put out a new model on Monday and said it's the biggest and most capable one the company has ever built.
GPT-5.6 Helps Two Teams Solve Quantum Crypto Puzzle Within Hours
Two arXiv submissions landed three hours apart, tackling the same unsolved problem in quantum cryptography, and both leaned on the same AI model to get there. MIT PhD student Seyoon Ragavan worked the problem alone.
Alibaba's Qwen3.8-Max Writes 7,600 Lines of Code in Five Days
Alibaba's Qwen team put out Qwen3.8-Max this week, a 2.4-trillion-parameter model with 95 billion active parameters per query, and the pitch is different from the usual chatbot upgrade.