Research & Benchmarks - Latest AI News & Updates
Academic AI research, performance benchmarks, scientific breakthroughs, and peer-reviewed studies advancing artificial intelligence frontiers.
853 articles
View complete article list
OpenAI reverses course on using Daybreak Blue cybersecurity data
OpenAI told members of its Trusted Access program this week that starting November 29, it will begin training models on Daybreak Blue cybersecurity session data.
AI Agent Told Wrong Answers Were Right: One Model Believed in Order Memory Test
A new paper argues that AI agents barely care whether their past attempts were labeled right or wrong.
New Infostealer Malware Uses AI to Automatically Target Victims
Cisco Talos researchers have found a new strain of infostealer malware that uses commercial AI models to sort through stolen data and flag which victims are worth the most.
Harness² Uses Eight Dormant Components to Improve AI Agent Performance
Google Research has a name for the thing nobody tunes when an AI agent stalls out: the harness. Not the model.
Altman empathizes with mathematicians but says OpenAI math onslaught won't stop
Sam Altman has a message for the mathematicians frustrated with OpenAI's methods: he gets it, but don't expect anything to change.
OpenAI doubles Pro usage limits after pushback as Anthropic competition heats
OpenAI raised the usage caps on its Pro subscription plans this week, a reversal that came only after subscribers complained loudly about planned cuts.
Experts adopt random AI agent opinions as their own, Anthropic study finds
Anthropic flipped a coin to settle an argument that didn't need settling, then watched 50 experts in causal inference fall for it anyway.
Rival AI Startup Announces USD 295 Million-Backed AI Agent for Code Verification
Chuyue Sun's experiment had a blunt result: an AI tool finished an engineering verification job in under half a day for less than $500. Three experienced engineers working by hand needed five months to do the same task.
Is the #1 AI Model Really #1 Across Eight Draws with Replacement?
A leaderboard puts a 71.8% score above a 70.0% score and calls it a ranking. Eight small open models, tested on the same 2,000 questions pulled from MMLU, say otherwise.
Block Unveils Dessa, a Full-Duplex Speech Model Trained Only on Speech
Block's AI research lab put out a speech model on October 23, 2026, that skips a step almost every voice assistant relies on.
Google DeepMind releases Gemini 4 Carbon; independent evaluations back gains
Google DeepMind put out Gemini 4 Carbon on November, just fourteen days after it shipped Argon, and the turnaround alone says something about the pace the company is running at right now.
DeepSeek-V4 Flash RL Runs on AMD Instinct MI308X GPUs
Three AMD Instinct MI308X nodes, 24 GPUs total, and one long-context reinforcement learning run on DeepSeek-V4 Flash.
New NVIDIA RTX Laptops Run AI Models Locally on CUDA
NVIDIA's RTX Spark laptops are now shipping from six manufacturers: ASUS, Dell, HP, Lenovo, Microsoft and MSI.
New Analysis Index Widens Gap Between Open and Closed AI Models
Kimi K3.1 Pro launched in fourth place on Artificial Analysis's Intelligence Index. Version 5 of that same index puts it ninth, behind eight closed models.
FEMA training may fade as AI emergency alerts approach and readiness lags
A county emergency manager in Alabama wanted residents to understand the threat from Hurricane Isaias earlier this month, so he turned to an AI chatbot to generate a graphic showing a "sting jet" of high winds bearing down on the coast.
Meta quietly launches Waddle World, a 16-bit AI pet journal app.
Meta has put out a pixelated pet simulator called Waddle World, and the pitch is stranger than the usual app store fare.
Google's AI Worlds Gain Voice in New Update
Google DeepMind's last world model, Genie 3, let people walk around an AI-generated scene for a few minutes before it fell apart. Genie 4, announced this week, adds something Genie 3 never had: a voice.
Oracle Adds Google TPUs, Cerebras to AI Compute Options
Oracle used its AI World conference in Las Vegas this week to announce it's widening the menu of chips available on Oracle Cloud Infrastructure, moving beyond its heavy bet on Nvidia.
Maya Research Releases Open-Weight Speech Model With 12 Languages and 20 Voices
Maya Research put out a new text-to-speech model this week called Maya 2 Native, and it covers 11 Indian languages plus English under an Apache 2.0 license.
Newest AI Quest Maps the Human Brain, Builds on Fruit Fly Brain Map
Google Research and the Stanford Accelerator for Learning released "Polar Peaks" on Tuesday, the third installment in their AI Quests series built for students ages 11 to 14.