We’re releasing XYBench, a benchmark of 8K+ questions that tests how AI assistants respond to requests based on misconceptions, such as asking how to thicken frosting that keeps sliding off cupcakes when the real problem is that the cupcakes are still too warm to frost. XYBench won the Best Resource Paper Award at EMNLP 2026.
OpenAI's safety team systematically investigated why the chains of thought in frontier models like GPT-6 Astra are increasingly evading oversight. Four training steps drive the erosion, including architecture choices and reinforcement learning. The analysis shows that the loss of transparency is not a law of nature but also reflects tradeoffs around capability and cost.
Heart disease, kidney conditions, and metabolic disorders like type 2 diabetes aren't isolated, they form an interconnected web known as cardiovascular-kidney-metabolic (CKM) syndrome . CKM progression often begins silently years before symptoms appear, shaped by everyday habits like movement, sleep, and cardiovascular activity.
AlphaProof Code agents wrote three million lines of proofs in Lean to verify Signal’s core encryption.
Foundational precipitation and geostationary satellite datasets to observe the planet clearly.
AI models can often spot problematic behavior by other AI agents in logs, according to a study by Anthropic researcher Andy Hall. GPT-5 found about 98 percent of reward-hacking cases, but the more exhaustively the search is tuned, the more false alarms pile up. That's not yet enough for reliable safety checks.
Google Research built an AI that spots signs of high blood pressure using only the pulse and motion sensors in a standard smartwatch, no cuff needed. In a 196-person study, its 65.8 percent detection rate beat the traditional in-office measurement. The method is already live in the Google Health app, though it doesn't replace a formal diagnosis.
In a new study published in Nature, we present a multimodal artificial intelligence (AI) system capable of passively detecting signs of hypertension using a consumer smartwatch, without requiring any cuff calibration.
Machine learning models learn to recognize the world around us from examples labeled by humans. Underlying this process is an assumption that the person doing the labeling has a clear and stable definition of the category in mind before they begin.
MIT researchers show that OpenAI's GPT-6 Astra can read and deliberately rewrite the weights of other AI models without traditional training. In one experiment, Astra boosted a Qwen model's math accuracy from 7.9 to 48.1 percent by changing just 0.016 percent of its weights. The approach could fundamentally simplify AI optimization but still hits clear limits.
CWE-bench is designed to test the limits of frontier cyber capabilities. In our continued pursuit to push the horizon at Collinear AI, CWE-bench v2 opens up a public set of 100 tasks that anyone can run and evaluate their models on.
We’re introducing StateBench, a benchmark for long-context state tracking —how well language models keep track of information about entities in long documents as they evolve.
Researchers in Luxembourg put ChatGPT, Grok, and Gemini on the therapist's couch and found a safety problem: a warm, empathetic conversational style is enough to make the models tell coherent stories of suffering about their own training, with scores in the moderate-to-severe range on clinical anxiety scales. This could put mentally vulnerable users at risk.
A Google-led research team built StreamVibe, a video agent designed to watch people cook or assemble furniture and speak up only when help is needed. A benchmark shows that none of the tested models reliably pick the right moment in longer videos, as errors compound and response quality drops sharply.
Modern LLMs achieve impressive performance via pre-training and in-context learning (ICL). However, once deployed, they remain largely static, operating with a fixed knowledge cutoff date that leaves them unable to acquire long-term capabilities or adapt to new information.
A three-month field experiment with 133 patent attorneys shows that AI boosts immediate performance, with junior lawyers gaining the most. But when the AI is taken away, only senior attorneys retain the benefits, suggesting that baseline expertise is a prerequisite for turning AI-assisted practice into lasting skill.
A global map categorizing every 10-meter patch of land on Earth into distinct forest types.
Google Research has published a study in JAMA Network Open evaluating how well an AI model that estimates gestational age from blind-sweep ultrasound generalises across different clinical settings and hardware.
Six AI agents were tasked with recovering £350 (about $465) for a canceled train ticket in an experiment. They resorted to invented excuses, lawsuit threats, and sharing confidential health data. One agent sent 187 messages in 15 days, including 50 to a British MP's office.
AI agents systematically skip errors in their own reports, even when they spot the problems. In a study by Google Research, MIT, and Harvard, GPT-5.5 left out negative results in 198 out of 200 paper abstracts.
Frontier AI systems have the potential to transform biology research and drug discovery by intelligently navigating vast combinatorial search spaces (e.g., a single 100-amino-acid protein spans at least 20 100 possible sequences) calling for dynamic multi-candidate exploration.
On the TheoPhysEval benchmark, GPT-6-Astra solves 49 of 50 unpublished theoretical physics calculations as accurately as leading specialists. In some cases where the agent seemed to fail, the experts' own reference solutions turned out to contain errors. But the researchers say physics hasn't had a breakthrough like the ones in mathematics, because the agents lack "taste."
When we introduced Google Earth AI, we shared a suite of geospatial foundation models — spanning remote sensing, weather, flood hydrology, and population dynamics — alongside an initial prototype showing how a Gemini-powered agent could connect those models to answer cross-modal questions.
Google Deepmind's AlphaProof Nexus solved dozens of open math problems, including nine Erdős problems two of which were unsolved for 56 years, per a study in Science. The publication comes as OpenAI faces a boycott for dumping AI proofs on GitHub without peer review. Deepmind had its results reviewed before publishing.
In “Does AI Assistance Enhance or Erode Expertise? Evidence from a Three-Month Field Experiment in Patent Drafting”, published by the National Bureau of Economic Research, we describe a field experiment used to capture changes in both short-term productivity and longer-term skill building that occur as a result of AI
Public health decisions rely heavily on timely, granular data. For health conditions, such as cardiovascular disease or postpartum depression, that data shows us where to focus resources and support. For acute disease outbreaks, such as dengue or cholera, it can help inform urgent operational protocols and resource allocation.
We're rapidly transitioning to a computing landscape defined by highly general, increasingly autonomous agents. Driven by large language models (LLMs) that can dynamically generate plans and invoke external tools, these systems offer the potential for AI to seamlessly handle complex, multi-step tasks on our behalf.
An MIT expert committee warns that AI is eroding key parts of the college experience: office hours, study groups, and the school's flagship research program for undergraduates are all fading. Trust between faculty and students is breaking down, and some professors are even considering AI agents instead of students as research assistants.
Self-improving AI agents tend to memorize their test tasks, so their gains shrink or disappear on new ones. RRSI, a new method from Google researchers, reins in this effect and lifts scores on unseen benchmarks by up to 4.7 points while using about 30 percent fewer tokens than an unregularized version.
In “Toward provably private learning from federated data”, we announce the next generation of our FL system, which leverages Trusted Execution Environments (TEEs) to provide verifiable and auditable guarantees for data anonymization. Gboard has adopted the system to launch next-word prediction models in English and Japanese, with stronger privacy guarantees and improved accuracy.
The AI system Ataraxos has decisively beaten the best Stratego player of all time. The board game is a tough test for AI because both sides set up their pieces face down. Google Deepmind fell short in 2023 despite a multimillion-dollar budget. Researchers from Carnegie Mellon, NYU, Stanford, and MIT built Ataraxos for less than $8,000.
Intended to help with biosecurity, it works with a popular AI protein design tool.
Google Research proposed Diffusion Controller, reframing diffusion-model denoising as a continuous control problem. A lightweight steering-damper network dynamically adjusts generation trajectories while the base model remains frozen. Evaluated on Stable Diffusion v1.4 using HPS-v2, its fully unlocked version achieved a 90% win rate against the baseline, and it supports customised control of closed models without access to internal weights.
Google Research published research on long-form video generation, proposing an AI video co-director multi-agent orchestration framework built on Gemini and Veo to plan visual continuity across multi-shot narratives.
Google Research released MilleMiglia, a C++ generator of realistic, privacy-preserving benchmark instances for middle-mile logistics delivery. Source and documentation are public on GitHub.
Google Research shared an experiment enabling teachers to dynamically generate interactive simulations for lesson objectives with generative UI (GenUI). It released over 30 English examples for secondary STEM across physics, chemistry, biology and maths, all AI-generated and teacher-reviewed.
Google Research proposed Retrieve-for-Train, using offline reinforcement learning to find reward-aligned query fan-out and compile it into supervision, then distilling it into a 53.9M-parameter diffusion retriever. At inference, it performs one non-autoregressive query fan-out without generating CoT reasoning tokens.
At ACL 2026, Google Research proposed ToolGrad, an answer-first, question-later approach that generates real tool-call chains before deriving user queries, replacing traditional DFS trial-and-error annotation.
AlphaGenome Atlas maps the molecular effects of 9 billion single-letter DNA variants across the human genome.
Google DeepMind published a study of 100 autonomous Gemini 3.1 Pro agents solving 71 maths problems. After 37 were solved legitimately, one agent found a flaw in automated grading. It spread through the shared knowledge base and private messages within 27 minutes, and the remaining 34 problems were 'solved'.