Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
A developer has created Genetic-Pareto Prompt Evolution (GEPA), a system that automates prompt optimization for AI agents by treating prompts as genomes that evolve through mutation and crossover. The method, featured in…
Fairplay announced plans to explore AI-powered personalization technologies to enhance user experience, improve content discovery, and increase platform accessibility, according to a June 3, 2026 press release. The compa…
AWS published a technical blog post detailing four private connectivity patterns for Amazon Bedrock AgentCore Gateway targets to keep traffic off the public internet for compliance and auditability. The patterns cover co…
Google Deepmind released Gemma 4 12B, an open AI model that natively processes text, images, and audio on laptops with just 16 GB of RAM. The model nearly matches the performance of its 26B counterpart on benchmarks whil…
A developer built Supply Chain Sense, an open-source AI inventory optimization suite that combines Google Gemini 2.5 Flash with classical Continuous Review Inventory Models to help small retailers manage stock. The syste…
Google Research has open-sourced its hydrology modeling framework on GitHub, allowing National Meteorological and Hydrological Services worldwide to integrate advanced AI-based flood forecasting into their own workflows.…
A major strategic pivot toward "Inference Economics" and hybrid AI infrastructure is reshaping how enterprises scale their AI systems, moving beyond the costly "cloud-only" model. By running lightweight, quantized models…
Nvidia announced that its Cosmos 3 model for physical AI ranks first on seven leaderboards for world generation, robot action policy, and industrial vision understanding. The company described Cosmos 3 as an "open omni m…
Fundamental’s NEXUS large tabular model is now available on Amazon SageMaker JumpStart, enabling enterprises to deploy a foundation model purpose-built for structured data prediction. The model generates deterministic, r…
DeepLearning.AI and Red Hat have released a free, intermediate-level course titled "Fast & Efficient LLM Inference with vLLM," taught by Red Hat Senior Developer Advocate Cedric Clyburn. The 1-hour 38-minute course teach…
Google DeepMind product managers Olivier Lacombe and Gus Martins announced Gemma 4 12B, a mid-sized open model under Apache 2.0 that runs multimodal and agentic tasks locally on laptops. The model bridges Google's edge-f…
Google released Gemma 4 12B, a dense multimodal model with a unified, encoder-free architecture designed to reduce latency and memory fragmentation for local AI applications. The model achieves strong performance in auto…
AWS has enabled SOCI snapshotter and index support on Deep Learning AMIs and Deep Learning Containers to reduce container cold start times for AI and ML workloads. The SOCI technology uses lazy loading to start container…
Google Cloud announced the general availability of its serverless Managed Service for Apache Spark runtime version 3.0, which introduces zero-setup onboarding and reduces startup times by 75%. The update automates IAM pe…
Amazon Web Services announced a new method for improving AI agent tool-calling accuracy using Supervised Fine-Tuning (SFT) and Direct Preference Optimization (DPO) on Amazon SageMaker AI. The approach trains small langua…
Researchers introduced Gated DeltaNet-2, a linear attention model that decouples the erase and write operations in recurrent state updates using separate channel-wise gates. The model outperforms Mamba-2, Gated DeltaNet,…
A new scientific paper argues that a mathematical theory of deep learning, termed "learning mechanics," is emerging, drawing parallels to physics by focusing on the dynamics of training and making falsifiable predictions…
Bazooka CIO Sankar Karuppasamy now reports directly to CEO Tony Jacobs following the company's 2023 separation from Topps via a $700 million sale to Apax Partners, reflecting a corporate trend where 65% of CIOs report to…
DeepSpeed has integrated the Muon Optimizer, a memory-efficient optimizer that uses a single momentum buffer and Newton-Schulz orthogonalization to improve training convergence, particularly for 2D weight matrices in tra…
NVIDIA Research presented three papers at the CVPR conference addressing challenges in physical AI, including a foundation model for robotic grasping, a faster reasoning model for autonomous vehicles, and a scalable trai…