A Google Research paper published at ICLR 2026, "On the Interpolation Effect of Score Smoothing in Diffusion Models," explains that score smoothing makes generation interpolate between training data, shedding light on where diffusion models' creativity comes from.
Hume released the Real World VoiceEQ benchmark to evaluate the human quality of voice AI interactions, covering ASR, TTS, S2S, and speech understanding, assessing over 40 leading closed- and open-source voice models across 15+ dimensions and 60+ metrics.
Google Research released SensorFM, a Large Sensor Foundation Model pretrained directly from unlabeled wearable data. The training data comes from 5 million users who consented to participate in the research, covering over one trillion minutes of multimodal sensor signals across more than 100 countries and over 20 Fitbit and Pixel Watch devices.
Google Research published a large-scale real-world study in Nature Cities, modifying Google Maps routing to divert less than 2% of trips away from congested roads across 10 US cities, in a six-month experiment using a city-level switchback design.
In Import AI issue 464, Jack Clark reports that Fable wrote a megakernel in CUDA achieving an 18.71X speedup on KernelBench-Mega.
Hugging Face launched the open-source benchmark ScarfBench (Self-Contained Application Refactoring Benchmark), which evaluates AI agents on enterprise Java cross-framework migration tasks across the Spring, Jakarta EE and Quarkus ecosystems.
Google Research's linear elastic caching approach, published at CIDR, models page eviction as a ski rental problem, using a lightweight decision tree model to predict TTLs for cached pages and dynamically resize caches. In Spanner production, it cut memory usage by 15.5% versus fixed-size caches while cache misses rose only 5.5%, lowering TCO by about 5% with only a 0.5% impact on actual I/O costs.
A Google Research paper at COLM 2026 found that enabling chain-of-thought reasoning allows Gemini-2.5 (Flash and Pro) and Qwen3-32B to recall simple single-hop facts they nearly always fail to answer when reasoning is turned off.
Import AI issue 462 rounds up multiple items: research from Oxford, the UK AISI, Stanford, and others, via 4 experiments and 18,978 conversations with 6,923 people, found that AI (strongest being Opus 4.1 and Opus 4.6) consistently surpasses expert humans at textual persuasion, raising 10.8 percentage points more in real donations than professional fundraisers; when AI is restricted to human speed and length, the advantage falls to insignificance.
Google released two dermatology AI studies, including a JAMA Dermatology survey of 2,345 people showing the AI-assisted group identified conditions with 23% accuracy, nearly triple the control search group's 8%, though the ability to judge next steps like seeking care did not significantly improve.
At AISTATS 2026, Google Research introduced Regularized f-Divergence Kernel Tests, a framework for auditing machine unlearning with greater sensitivity and flexibility.
Google DeepMind has released results of a pre-registered trial in Sierra Leone: after eight weeks of using Guided Learning, students' math scores improved by +0.258 standard deviations versus the control group, equivalent to about 1.2 to 1.7 years of typical learning progress; classes where teachers integrated Gemini into about half of lessons, reaching the 12-hour target, progressed even more, about 1.8 to 2.5 years.
Why it matters: Google DeepMind shares teaching details and quantified results from the pre-registered Sierra Leone trial, and has opened a teacher training guide and RCT playbook so readers can learn a replicable research approach.
Google published the PHRM research system in Nature, which uses facial video captured by the front camera after face unlock to estimate heart rate and daily resting heart rate (RHR) in the background via deep learning.
Why it matters: The study covers nearly 700 participants of diverse skin tones and was validated in real-life scenarios, giving readers accuracy data for passive camera-based heart rate measurement plus open data access.
Import AI issue 459 reports: researchers from the University of Virginia, Anthropic, and the Bank of Canada estimate the US AI economy at roughly $250 billion nominal in 2025, with quality-adjusted real output growing about 2,600% a year, yet hard to see in GDP because of falling prices and the high share of inference. The UK AI Security Institute argues that using AI to oversee training processes to automate alignment research is not a panacea and is harder in practice than expected.
SentinelOne researchers dissected a virus from roughly 20 years ago, fast16.sys, which tampered with in-memory computation results of high-precision engineering software (such as LS-DYNA 970, PKPM, and MOHID), potentially enabling sabotage related to weapons development — reminiscent of the sophon interference in The Three-Body Problem.
A team led by Stanford School of Medicine geneticist Gary Peltz published research in Advanced Science using Google DeepMind's Co-Scientist to screen existing drug literature for candidates that can be repurposed to treat liver fibrosis.
Google DeepMind released the Decoupled DiLoCo distributed training architecture, which splits large model training into asynchronously communicating compute islands, isolating hardware failures and sharply reducing bandwidth requirements.
Google Research has published the ICLR paper ReasoningBank, proposing a framework that distills high-level reasoning memory from an agent's successful and failed experiences, distinguishing it from Synapse and AWM, which only record trajectories or summarize successful workflows.
Berkeley AI Research proposes GRASP, a gradient-based world model planner that makes long-horizon planning more practical. Three core improvements: lifting trajectories into virtual states so optimization runs in parallel over time, injecting randomness directly into state iterations to support exploration, and reshaping gradients to avoid traversing fragile state-input gradients of high-dimensional vision models.
Google Research has published a paper in TMLR introducing Simula, a reasoning-first synthetic data generation framework. Drawing on mechanism design, it breaks data generation into four controllable axes — global diversity, local diversity, complication and dual-critic quality checks — without needing seed data.
Google Research has proposed MoGen, a neuron morphology generation model based on PointInfinity point cloud flow matching. Trained on 1,795 human-verified mouse axons, it generates synthetic neuron shapes to expand training data for the PATHFINDER reconstruction model.
Google Research released the ConvApparel dataset, containing over 4,000 human-AI multi-turn conversations spanning nearly 15,000 turns, to quantify the "realism gap" in LLM user simulators.
Google Research proposes a framework for evaluating how well the behavioral dispositions of 25 LLMs align with humans, adapting standardized psychological questionnaires into situational judgment tests (SJTs), with human preference distributions provided by 550 participants.
In the paper 'Forest vs Tree: The (N, K) Trade-off in Reproducible ML Evaluation,' Google Research studies the trade-off in AI evaluation between the number of annotations per item and the number of items annotated, and open-sources a simulator based on real datasets.
Google Quantum AI published a white paper updating quantum resource estimates for breaking the 256-bit elliptic curve discrete logarithm problem (ECDLP-256): two circuits use under 1,200 logical qubits with 90 million Toffoli gates and under 1,450 logical qubits with 70 million Toffoli gates, cutting physical qubit requirements by roughly 20x versus previous estimates.
Why it matters: The Google white paper offers lower quantum resource estimates for breaking ECDLP-256 along with a responsible zero-knowledge-proof disclosure approach, letting the crypto community gauge the urgency of migrating to PQC.
Google Research introduced TurboQuant, a compression algorithm (to be published at ICLR 2026) that combines Quantized Johnson-Lindenstrauss (QJL) and PolarQuant (to be published at AISTATS 2026) to achieve zero-memory-overhead vector compression, compressing models and KV cache without accuracy loss.
Google Research, under its Earth AI program, introduced S2Vec, a self-supervised framework that learns universal embeddings of the built environment. It uses S2 Geometry to divide the surface into multi-resolution cells, rasterizes features such as buildings and roads into layered images, and applies masked autoencoding to learn location features without human labels.
AIMS, a collaboration between Google Research and the UK's NHS, has published two papers in Nature Cancer evaluating an AI breast cancer detection system.
Why it matters: The original articles give the specific results and figures of the two Nature Cancer studies, helping readers understand how AI performs in dual-reading breast screening workflows and where it falls short.
Google Research, with Cornell and Harvard, published a PNAS paper in which 12 high-temperature superconductivity experts wrote 67 questions and blind-reviewed answers from 6 LLMs.
Berkeley AI Research introduces SPEX and ProxySPEX, two algorithms that use structural assumptions such as sparsity, low degree, and hierarchy to turn interaction discovery in ablation attribution into a sparse recovery problem, identifying the key interactions that drive LLM outputs.
Google Research partnered with Beth Israel Deaconess Medical Center on a prospective single-center feasibility study in which AMIE gathered medical histories before primary care visits for 100 adult patients, with physicians supervising in real time via video.
Why it matters: The original presents prospective feasibility data and safety results for AMIE in a real primary care setting, showing the current evidence for conversational medical AI moving toward clinical use.
Berkeley AI Research proposes a NeurIPS 2025 framework that evaluates and optimizes imaging systems directly on information content, using a mutual information metric to unify factors like resolution and noise. The method needs only noise measurements and a noise model to estimate information, and provides an upper bound on true information. Across color photography, radio astronomy, lensless imaging, and microscopy, the metric predicts decoder performance and its optimized results match SOTA end-to-end methods with lower memory and compute and no task-specific decoder.
Berkeley AI Research proposes an off-policy RL value-learning algorithm based on divide and conquer that no longer uses temporal difference (TD) learning, whose Bellman recursion error accumulation makes it hard to scale to long-horizon tasks.
Berkeley AI Research proposes a quantitative theory of the word2vec learning process: under several mild approximations, its learning problem reduces to unweighted least-squares matrix factorization, the gradient flow dynamics have a closed-form solution, and the final embeddings are equivalent to PCA on the target matrix M*.