#google research
Google Research: 10 AI articles covering google research news, analysis, and research
Articles
Google Research's RRSI Guide: Building Self-Improving AI AgentsNEWβ9
RRSI lets LLM agents rewrite their own harness without overfitting. Learn its selection rules, noise band, edit budget, and how to audit it.
EmTech Future 2026: When AI Meets Everythingβ8
Discover highlights from EmTech Future 2026, where MIT Technology Review explores how AI is reshaping biology, infrastructure, manufacturing, science, and more.
Google Research Moves Federated Learning Into TEEs: Gboard Nowβ9
Google Research brings federated learning into TEEs, enabling externally verifiable central differential privacy. Gboard now trains next-word prediction and Sma...
Google Research Open-Sources RRSI: AI Agents That Improve Theirβ8
Google Research open-sources RRSI, a framework that lets AI agents recursively improve their own prompts, tools, and memory without overfitting or changing mode...
Google Research Unveils AI Video Co-Director: 4 Agenticβ8
Google Research's AI video co-director uses four agentic frameworksβCo-Director, CANVAS, and moreβto generate coherent, minutes-long videos and fix identity dri...
Google to Launch AI Satellite Into Space: A New Frontier forβ9
Google is launching an AI-powered satellite to test orbital AI processing, aiming to cut latency, save bandwidth, and build resilient space-based cloud computin...
Google Research Unveils Retrieve-for-Train (R4T): An RL-Compiledβ8
Google Research's Retrieve-for-Train (R4T) uses RL and a diffusion retriever to generate diverse query fan-outs 12β20Γ faster, avoiding paraphrastic collapse in...
Google Research Releases ToolGrad: Answer-First Framework Hitsβ8
Google Research's ToolGrad flips tool-use data generation to answer-first, achieving a 99.8% pass rate and letting Gemma-3 match frontier models with just 500 s...
Google Research Introduces GlucoFM: A 0.72M-Parameter Dualβ7
Google's GlucoFM: a 0.72M-parameter dual-stream AI model for continuous glucose monitoring, boosting accuracy in diabetes tasks with efficient training.
Google Research Introduces ME-POIs: A Mobility-Informedβ7
Google's ME-POIs framework enhances place embeddings by integrating human movement patterns, boosting map tasks like visit intent and busyness prediction.