Microsoft Research has developed an end-to-end machine learning pipeline that uses solar wind observations at L1, forecasts of the AE and Dst indices, and local geological conductivity data to estimate local space weather risks 30 to 60 minutes ahead for 66,935 substations in the contiguous United States.
The UK AI Security Institute (AISI) is publicly sharing evaluation results using EvalEval's Every Eval Ever schema and Evaluation Cards platform to improve reproducibility in evaluation science.
Microsoft published RetroChimera, a retrosynthesis prediction framework, in Nature and open-sourced its implementation and weights. It combines the Transformer model R-SMILES 2 with GNN-based NeuralLoc, using a learned ensemble to rerank predictions. Chemists preferred its single-step reaction predictions in blind testing.
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.
Google DeepMind launched AlphaGenome Atlas, a platform containing predicted effects for 9 billion single-nucleotide variants across the human genome. At 1 PB, it is over 30 times the size of the AlphaFold Database.
Google Research evaluated cross-population transfer of polygenic risk scores (PRS) on eight clinical measures using European UK Biobank data and nearly 200,000 Japanese participants from Biobank Japan. At target-population samples of 15,000 or more, target-specific models outperformed mixed European-data training, while highly genetically correlated traits continued to benefit from European data until target samples reached 25,000–40,000 or more.
Google Research partnered with HHMI Janelia, the University of Cambridge and others to publish in Cell a complete connectome of a male fruit fly’s brain and central nervous system. It contains over 166,000 neurons and 125 million synaptic connections, making it the largest brain map to date by neuron count.
Google Research proposed MAPL-EMIT, a Swin-S Transformer-based deep-learning framework automatically detecting methane plumes, predicting enhancements and locating them in NASA EMIT hyperspectral satellite data, with 84% recall on expert-labelled plumes.
Google Research introduced the experimental Planetary Prediction Engine (PPE) under Google Earth AI. From a natural language query, it autonomously discovers geospatial data, engineers features, trains and evaluates models and produces reports, compressing weeks of manual data engineering into minutes.
Google released self-supervised GlucoFM, separating slow glucose trends from short-term fluctuations. Pretrained on 109,066 hours of unlabelled CGM data, it averaged 5.8 percentage points higher PR-AUC than the best GluFormer variant across 14 evaluations on seven clinical tasks in four cohorts, and achieved the lowest post-meal response prediction MAE.
Multiverse Computing published a paper proposing Quantization-Aware Healing (QAH). After compressing GPT-OSS 120B to 60B parameters and quantising it to MXFP4, the method distils directly from the original uncompressed model rather than a reconstructed bfloat16 checkpoint.
METR research shows AI's effects on scientific discovery are uneven: cybersecurity vulnerability reporting accelerated sharply in 2026 compared with 2025, mathematics accelerated only modestly, and algorithmic progress in AI research itself showed no significant acceleration. A multi-university team proposed SPADE, alternating LLM generation of executable training environments with solving them. Qwen3-30B-A3B averaged 58.3 on the game-environment suite, 8.1 above baseline.
Google introduced Biomarker Discovery Framework, a human-supervised multi-agent system organising candidate biomarker prioritisation into iterative research cycles. Across 9,279 participant observations in three cohorts, it automatically identified 41 candidate digital mental health biomarkers and 25 metabolic candidates, including an association between sleep-duration variability and PHQ-8 severity (ρ = 0.252).
Google Research proposed Mobility-Embedded POIs (ME-POIs), combining anonymised mobility patterns with text descriptions to generate place embeddings encoding both identity and dynamic function.
Microsoft Research released deep-learning DFT functional Skala 1.1, trained on 2.5 times more data than the previous public version. It ranks first in 32 of GMTKN55's 55 categories with 2.8 kcal/mol weighted average error. Available in CP2K and being integrated into Psi4, FHI-aims, ORCA and VASP, it is tracked by a new living performance benchmark.
Import AI 469 introduces DiG-bench (Discovery in Games), with 70 games whose hidden rules and goals must be discovered through interaction. Opus 5 and Fable 5 with Claude Code performed best; only they completed some Tier 7 tasks, while humans achieved 100% completion.
Google Research's knowledge profiling framework evaluated 13 LLMs on WikiProfile's 2,150 Wikipedia facts. Gemini-3-Pro and GPT-5 encoded 95–98% but failed direct recall of 26–34%, and still failed 11–12% with thinking enabled. It argues frontier factual errors arise more from knowledge accessibility than absence, shifting the bottleneck from acquisition to use.
Microsoft Research released CARE-X, a unified chest X-ray vision-language research model for report generation and structured prediction. It rewards clinical correctness using multitask reinforcement learning (DAPO). Generation and dual-inference modes cover lesion presence and negation, localisation, multilabel classification, catheter and tube malposition detection, and localisation of 29 anatomical regions.
Multiverse Computing published the paper Efficient Knowledge Distillation for LLMs: Offline Top-K Logits and a Fused Chunked KL Loss.
Microsoft Research proposed Echoverse, constructing twelve training worlds for computer-use agents: ten deep domain worlds and two capability worlds. Code, data and scorers for four worlds are open-sourced.
Berkeley AI Research proposed ABBEL, isolating and supervising LLM summaries as natural-language belief states and using autoencoder-inspired reconstructive belief scoring as an auxiliary RL task. On CollabBench, ABBEL-rec-BG narrowed the performance gap to the full-context model by around 50%, reduced training steps from 100 to 50 and used a shorter peak token length.
A Google Research ICLR 2026 paper traces diffusion model creativity to score smoothing from regularisation such as weight decay. Smoother learned score functions let denoising interpolate between training points, generating new samples rather than simply memorising them.
Google Research and Google DeepMind proposed SensorFM, a large sensor foundation model learning directly from unlabelled wearable data. Pretraining uses over 1 trillion minutes of multimodal sensor signals from 5 million consenting participants across more than 100 countries and over 20 Fitbit and Pixel Watch devices.
Hugging Face proposed DiScoFormer (Density and Score Transformer), estimating a distribution's density and score simultaneously in one forward pass from a set of data points, without retraining for new distributions.
Google Research released vector data converting Farmscapes 2020 high-resolution raster maps into usable inventories of hedgerows, stone walls and coppices across over 130,000 square kilometres of the UK. The framework fine-tunes an RSF Vision-Transformer pretrained on over 300 million global satellite images with around 247 square kilometres of labelled data. Polsby–Popper compactness distinguishes woodland, clusters of trees and hedgerows, with a threshold below 0.5 for linear features.
Google DeepMind published results from a preregistered randomised controlled trial in Sierra Leone. Over eight weeks, students using Guided Learning improved maths scores by 0.258 standard deviations over controls, equivalent to around 1.2–1.7 years of conventional progress.
Google released Empirical Research Assistance (ERA), a research tool using Gemini to write and optimise scientific code. Its paper was published in Nature today, and the tool is available to scientists worldwide as part of Gemini for Science.
Import AI 457 focuses on three studies. SentinelOne analysed fast16.sys, malware from over 20 years ago that sabotages engineering and physics simulations by altering in-memory results from precision software such as LS-DYNA 970, PKPM and MOHID.
Calico Life Sciences' Matt Onsum and Katherine Labbé use Google DeepMind Co-Scientist to integrate scattered ageing biology findings and generate testable hypotheses.
Gary Peltz's Stanford medical team used Google Co-Scientist to screen drugs. Two of three suggested candidates blocked fibrosis and promoted regeneration in living human liver cells, while both Peltz-selected drugs failed. Cancer drug vorinostat blocked 91% of damage responses driving scarring. The work appeared in Advanced Science.
Google DeepMind published a paper proposing Decoupled DiLoCo, splitting large-scale training into decoupled compute islands that communicate through asynchronous data streams, isolating local hardware failures.
Google proposed Simula, a reasoning-first synthetic data framework reframing dataset construction as mechanism design. It builds datasets from scratch without seed data through global diversification, local diversification, complexification and dual-critic quality checks.
Google Research proposed MoGen (Neuronal Morphology Generation), using PointInfinity point-cloud flow matching to generate synthetic neuronal shapes and supplement PATHFINDER reconstruction training. It reduced reconstruction error by 4.4% on held-out mouse axons, mainly by reducing merge errors.
Google Research studied trade-offs between evaluation item counts and annotators per item in Forest vs Tree: The (N, K) Trade-off in Reproducible ML Evaluation. Common counts of one, three or five annotators are often insufficient; reflecting human disagreement frequently requires more than ten.
S2Vec uses S2 Geometry partitions and rasterised features to turn buildings and roads into multilayer images, learning general embeddings with masked autoencoding (MAE). It beats SATCLIP and GEOCLIP on zero-shot geographic socioeconomic extrapolation such as US population density and median income, but needs improvement on tree cover and elevation.
Berkeley AI Research proposed SPEX and ProxySPEX to identify key interactions driving LLM outputs at scale in feature, data and model-component attribution. SPEX turns interaction search into sparse recovery using sparsity and low order; ProxySPEX exploits hierarchy to match SPEX with roughly ten times fewer ablations.
Google Research released WAXAL, a large open speech dataset initially covering 27 sub-Saharan African languages spoken by more than 100 million people, under CC-BY-4.0.
Berkeley AI Research proposed a mutual-information framework to evaluate and optimise imaging from noisy measurements and noise models. A NeurIPS 2025 paper validates decoder performance predictions for colour photography, radio astronomy, lensless imaging and microscopy. Designs match end-to-end state-of-the-art approaches with less memory and compute, without task-specific decoder design.
Berkeley AI researchers proposed the first theory quantitatively predicting word2vec learning. Under realistic training conditions, it reduces to unweighted least-squares matrix factorisation with a closed-form gradient-flow solution, and the resulting representation is equivalent to PCA of target matrix M*. From small initialisation, word2vec learns orthogonal linear subspaces in discrete steps, increasing embedding rank and reducing loss. These features are M*'s leading eigenvectors and can be calculated in advance from corpus statistics and hyperparameters.