Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
A developer published a tutorial showing how to build a real-time heart-rate-variability outlier detector using a bidirectional LSTM in TensorFlow/Keras, feeding raw R-R interval timestamps from wearables such as the App…
Luke Salamone trained LLM reasoning agents at 0.8B, 2B and 9B parameters to play chess by calling tools that track the board state, enabling strong performance with relatively few parameters. The demo code is available o…
A developer built Spotted, a browser-based field guide PWA that identifies birds, lizards, insects, trees and plants from a single photo using an 8-bit ONNX build of OpenAI's CLIP ViT-B/32 as a zero-shot classifier, then…
A developer built a self-supervised learning pipeline for coastal climate resilience planning that mines temporal patterns from extremely sparse sensor data, targeting Small Island Developing States where in-situ buoy me…
Google Research released TabFM, a foundation model for tabular data that predicts missing labels in a single forward pass by treating labeled and unlabeled rows as an in-context prompt, eliminating per-dataset training a…
Google Research's TabFM, a tabular foundation model trained entirely on synthetic data generated from random causal graphs, predicts held-out cells in a single inference pass without training on the target dataset, and o…
A paper published in October 2026 by Jianhua Li, Yun Xing, Girisha De Silva, Ryan Flint, Kai Yuan and Sean Suchter presents Minions, a dynamic and distributed virtual GPU cloud that aggregates idle Apple Silicon GPUs acr…
A fine-tuned RoBERTa encoder reached 0.78 macro-F1 on a 1,232-item test split of the author's tldr_news classification taxonomy, 30 to 40 points above every zero-shot system tested, including Anthropic's Claude Haiku 5.5…
A developer built VCIS-Bench (Visual Constraint & Instruction Simulation), a synthetic 7×7 grid puzzle benchmark testing whether vision-language models can perform step-by-step spatial verification, and evaluated Claude …
Sakana AI and UC Berkeley developed MASS, a method that uses one model to organize research agents, judge their results, and train on selected team runs, according to Sakana AI's MASS project page. The method produced ga…
Coinbase reported that a fine-tuned version of Alibaba's Qwen3.5-9B model outperformed frontier models including Anthropic's Opus 4.5 on its proprietary Onramp fraud benchmark, cutting fraudulent transactions by 30% and …
A developer built MicroCast, an offline, zero-shot mountain squall and microclimate predictor for backcountry hiking that runs locally without cellular reception. The tool uses Prior Labs' TabPFN tabular foundation model…
An AI model developed by researchers at NYU Langone Health and its Perlmutter Cancer Center estimated a woman's five-year breast cancer risk by analyzing multiple years of 3D mammograms rather than only the latest scan, …
A developer built BirdEcho, an open-source web app that identifies bird species from recorded audio using the BirdNET+ V3.0 model running entirely in the browser via ONNX Runtime Web and WebAssembly. To keep the interfac…
A developer built Ink, a system that lets AI agents earn local execution authority for repeated bounded decisions only when the lower confidence bound on verified accuracy clears a site's required threshold. In a frozen …
Epoch AI's InnovationEval benchmark shows that current AI models lag human researchers at replicating algorithmic breakthroughs, indicating a measurable capability gap in novel problem-solving. The benchmark scored 89 on…
A developer built Grassday, an offline-first AI companion that predicts a user's optimal two-hour outdoor window each day and generates a printable field card so they can leave their phone behind. The tool fits a local T…
A developer built Maati, an open-source web app and optional Windows desktop companion that tracks screen time and nudges users outside with small, time-of-day-appropriate missions. The stack pairs a Next.js/TypeScript/T…
Researchers from Sungkyunkwan University, Samsung Research, and KAIST presented the Unlearning Depth Score (UDS), a mechanistic metric that quantifies how deeply a model has unlearned target knowledge by measuring how mu…
NumPy 2.5, released in June, rewrote the searchsorted function in C++ to make batched binary searches up to 25x faster, according to the Scientific Python blog post detailing the optimization. The speedup applies to arra…