Google/DeepMind

Latest AI news, models and releases from Google/DeepMind. ['A2UI v0.9', 'AI Co-Scientist', 'AI Evaluator', 'AI Mode', 'AI Overviews', 'Alexa', 'AlphaEvolve', 'AlphaFold', 'AlphaGenome', 'AlphaQubit', 'Antigravity', 'Auto frame', 'BERT', 'BioBERT', 'Chinchilla', 'ClinicalBERT', 'Computational Discovery', 'Confidential GKE Nodes', 'Co-Scientist', 'Empirical Research Assistance', 'Era', 'ERA', 'Executive LLM', 'Farmscapes 2020', 'Flood Hub', 'frontier AI', 'Frozen v2', 'FunctionGemma', 'Gemini', 'Gemini 1.5 Flash', 'Gemini-1.5-pro', 'Gemini 1.5 Pro', 'Gemini 2.0 Flash', 'Gemini 2.0 Flash-Lite', 'Gemini 2.5', 'Gemini 2.5 Flash', 'Gemini-2.5 Flash', 'Gemini-2.5-Flash', 'gemini-2.5-flash-image', 'gemini-2.5-flash-lite']

Official site ↗

Agents 🇺🇸

Google Introduces Two AI Agents for Scientific Research: PaperVizAgent for Illustrations and ScholarPeer for Reviewing

Google Research has unveiled PaperVizAgent and ScholarPeer, two AI agents designed to assist academic research. PaperVizAgent generates publication-ready figures from text, while ScholarPeer automates rigorous peer review with web-grounded literature analysis.

Google/DeepMind

Google Research27.07 · 16:04

Research 🇺🇸

ConvApparel: Measuring and Closing the Realism Gap in LLM-Based User Simulators

Google Research introduces ConvApparel, a new human-AI conversation dataset and three-pillar evaluation framework to quantify and reduce the 'realism gap' in LLM-based user simulators. The framework includes population-level statistics, human-likeness scoring, and counterfactual validation, applied to simulators built with Gemini models. Results show data-driven simulators (ICL, SFT) outperform prompt-based ones, but even the best models still exhibit subtle synthetic artifacts.

Google/DeepMind

Google Research27.07 · 16:04

Show more ↓