News
Improving the academic workflow: Introducing two AI agents for better figures and peer review
21+ hour, 56+ min ago (345+ words) Jinsung Yoon, Research Scientist, and Tomas Pfister, Director, Google Cloud Introducing two AI agents to streamline academic research. These include: Paper Viz Agent, a visualizer agent for drawing academic figures, and Scholar Peer, a reviewer agent that automatically and rigorously…...
Evaluating alignment of behavioral dispositions in LLMs
6+ day, 2+ hour ago (446+ words) Our objective is to build upon such psychological questionnaires, but directly applying them to LLMs presents technical challenges, as LLM outputs are sensitive to prompt phrasing and distribution shifts. Consequently, dispositions "claimed" by LLMs within a self-report format are not…...
Building better AI benchmarks: How many raters are enough?
1+ week, 2+ day ago (563+ words) Flip Korn and Chris Welty, Research Scientists, Google Research We introduce an evaluation framework for ML models, based on "gold" ratings data, that optimizes the trade-off between the number of items and raters per item, providing a roadmap for building…...
Safeguarding cryptocurrency by disclosing quantum vulnerabilities responsibly
1+ week, 2+ day ago (1041+ words) research. google Safeguarding cryptocurrency by disclosing quantum vulnerabilities responsibly Ryan Babbush, Director of Research, Quantum Algorithms, and Hartmut Neven, VP of Engineering, Google Quantum AI, Google Research We're exploring a new model for how to elucidate the code breaking capabilities…...
Vibe Coding XR: Accelerating AI + XR prototyping with XR Blocks and Gemini
2+ week, 22+ hour ago (747+ words) Ruofei Du, Interactive Perception & Graphics Lead, and Benjamin Hersh, Product Manager, Google XR Vibe Coding XR is a rapid prototyping workflow that empowers Gemini Canvas with the open-source XR Blocks framework to translate user prompts into fully interactive, physics-aware Web…...
Turbo Quant: Redefining AI efficiency with extreme compression
2+ week, 1+ day ago (347+ words) Amir Zandieh, Research Scientist, and Vahab Mirrokni, VP and Google Fellow, Google Research We introduce a set of advanced theoretically grounded quantization algorithms that enable massive compression for large language models and vector search engines. Turbo Quant is a compression…...
Mapping the modern world: How S2 Vec learns the language of our cities
2+ week, 2+ day ago (474+ words) To address this challenge, S2 Vec uses a two-step process to rasterize the world: S2 Vec rasterizes images to learn embeddings of the built environment. The MAE process systematically shows the model a "patch" of the built environment while hiding (masking) certain…...
Improving breast cancer screening workflows with machine learning
3+ week, 1+ day ago (258+ words) Lihong Xi, Senior Technical Program Manager, and Daniel Golden, Engineering Manager, Google Research The first study was split into two phases. In the first phase, we conducted a large-scale multi-center retrospective evaluation of the standalone performance of the AI system....
Testing LLMs on superconductivity research questions
3+ week, 3+ day ago (397+ words) Can LLMs become expert-level research partners in modern physics? Using high-temperature superconductivity as a case study, physicists tested six LLMs with challenging questions and graded the responses. Several other groups across Google are also exploring AI to advance scientific research:…...
Introducing Groundsource: Turning news reports into data with Gemini
4+ week, 4+ hour ago (346+ words) Oleg Zlydenko, Software Engineer, Rotem Mayo, Software Engineer, and Deborah Cohen, Research Scientist, Google Research Today, we're introducing Groundsource, a new scalable methodology that leverages Gemini to transform unstructured global news into actionable, historical data. Our first, open-access Groundsource dataset…...