Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
Swiggy developed a real-time machine learning ranking system for search autocomplete that replaces hand-tuned heuristics with a learned ranking model running directly inside OpenSearch, operating under strict latency con…
The Open Agent Leaderboard is a new open benchmark designed to evaluate the performance and cost of full AI agent systems—including their tools, planning, and error recovery—rather than just the underlying models. It mea…
SFHformer, an image restoration framework that integrates Fast Fourier Transform (FFT) with Transformer architecture to address various image degradations like dehazing, deraining, and deblurring. By analyzing degradatio…
Findings from Project Glasswing, where security-focused LLMs, particularly Anthropic's Mythos Preview, were tested against over fifty internal code repositories. It highlights Mythos Preview's advanced capabilities in ex…
A 70-billion parameter model can cost less per useful token than an 8-billion parameter model in production inference, because the real cost driver is GPU utilization rather than parameter count. Teams that switch to sma…
Vicki Boykis added tags to her static blog using BERTopic and LLMs including Claude Code and Deepseek, completing the project in 6-10 hours. She argues that LLMs excel at compression tasks like topic modeling and labelin…
Researchers applied the Bass diffusion model to estimate the lifecycle of open-source projects by fitting monthly developer contributions and extrapolating future activity. The model's use of the hyperbolic secant functi…
Project Genie, a general-purpose world model from Google, now integrates Google Street View imagery to ground AI-generated virtual environments in real-world locations. This new capability allows users to select a real p…
The article explores morphogenesis, the self-organizing process by which a single cell develops into a complex multicellular organism, highlighting its robustness and regenerative capabilities. It discusses the challenge…
PyTorch introduced a "fake export" method that allows developers to verify the exportability of large deep learning models using `torch.export` APIs without requiring actual GPU memory. The approach uses fake tensors wit…
Researchers introduced Iterative Group Relative Policy Optimization (iGRPO), a two-stage reinforcement learning method that improves large language model reasoning by having the model generate and refine its own best dra…
Researchers have developed RLP, an information-driven reinforcement pretraining objective that integrates exploration and chain-of-thought reasoning into the pretraining phase of large language models, rather than reserv…
A new study published in Science Advances details electronic rings that wirelessly transmit hand motion data to an AI system capable of translating multiple sign languages into text. The seven-ring system, developed by r…
Google's Gemma 4, DeepSeek V4, and other new open-weight large language models incorporate architecture changes like KV sharing, compressed convolutional attention, and layer-wise attention budgeting to reduce long-conte…
Researchers at Calico Life Sciences, including AI/ML head Matt Onsum and principal scientist Katherine Labbé, are using an AI tool called Co-Scientist to analyze the complex biology of aging and generate testable hypothe…
In October 2025, the National Hurricane Center (NHC) used Google DeepMind's AI model, WeatherNext, to predict Hurricane Melissa's historic Category 5 landfall in Jamaica five days in advance, marking the first time a sto…
The article is a blog post about the bigfin squid that also invites discussion of unaddressed security news. The comment section features a lengthy argument by Clive Robinson that the rise of AI and corporate cost-cuttin…
DeepSeek-V4-Flash, a new local AI model, now makes it practical for engineers to experiment with "steering" — a technique that manipulates a model's internal activations to guide its outputs. The stripped-down version, D…
Amazon Web Services has introduced document-level access control list (ACL) support for Amazon Simple Storage Service (Amazon S3) knowledge bases in Amazon Quick, allowing organizations to restrict sensitive documents to…
Amazon researchers have developed a new scaling law that connects large language model architecture directly to inference efficiency, enabling faster text generation without sacrificing accuracy. The framework, presented…