Machine Learning News

NVIDIA released a workflow for converting FP8-quantized CLIP model checkpoints into TensorRT engines, enabling faster inference and higher GPU throughput for production deployment. The process involves exporting the quan…

Google released Gemini 3.5 Live Translate, a streaming speech-to-speech audio model that translates over 70 languages in real time. The model processes audio continuously rather than waiting for a speaker to finish, pres…

NVIDIA introduced an automated, AI-driven research loop called Auto-FL within its NVIDIA FLARE platform to help researchers test and optimize federated learning strategies more efficiently. The system uses bounded AI age…

Google Cloud announced the general availability of activity insights within Storage Insights datasets, providing administrators with operational visibility into how storage objects are accessed, moved, and modified. The …

Google Cloud's GKE Inference Gateway delivers up to 92.8% shorter wait times and 62.6% lower inter-token latency compared to the next leading managed Kubernetes service, according to an independent benchmark report. The …

Google released Gemini 3.5 Live Translate, a new audio model for live speech-to-speech translation that automatically detects over 70 languages and generates continuous, natural-sounding translated speech while preservin…

Microsoft researcher Lorin Crawford and his team published a study in Nature Methods that uses AI to analyze how individual cancer cells behave and interact with their environment, aiming to explain why patients with sim…

NVIDIA has introduced a new workflow using agent skills and Nemotron Speech to accelerate the evaluation of clinical automatic speech recognition (ASR) models, addressing the challenge of accurately recognizing rare medi…

Hugging Face hosts over 90,000 text-to-image models, but only seven are worth using in 2026 for free, high-quality image generation. The list includes FLUX.1 Schnell, which is released under the permissive Apache 2.0 lic…

A new deep learning model can accurately track glacier calving fronts with minimal manual input, reducing error margins from over a kilometer to under 70 meters per glacier. Researchers at Friedrich-Alexander University …

GPU clouds face a fundamental tension as they attempt to carve expensive hardware into shareable units for enterprise AI workloads, but the hardware was never designed for safe multitenant use, fast fault recovery, or cl…

Hugging Face launched a new GitHub Actions integration that allows developers to run CI jobs on Hugging Face's infrastructure instead of GitHub's default runners. The setup uses a dispatcher Space that receives GitHub we…

AI scraping uses artificial intelligence to extract data from websites by understanding content meaning rather than relying on fixed structural rules, allowing it to adapt to site changes and handle JavaScript-rendered c…

SigNoz has released a guide for setting up observability and monitoring for Mistral AI applications using OpenTelemetry, enabling developers to export traces, logs, and metrics to the SigNoz platform. The integration all…

Neuronpedia launched HeadVis, a tool for exploring attention heads in collaboration with Anthropic, now available for 37 models with over 36,000 attention head dashboards. The platform also added new SAEs and NLAs, inclu…

A PyCon US 2026 talk presented findings from an experiment on type checking in agentic workflows. The study found that type checking improved agent success rates from 80% to 84% only in well-typed codebases, while low ty…

SigNoz released a new integration guide for monitoring Hugging Face AI applications using OpenTelemetry, enabling developers to export traces, logs, and metrics to the SigNoz observability platform. The guide provides st…

Alibaba Cloud has released a guide for setting up observability and monitoring for its Qwen large language model using OpenTelemetry and SigNoz. The integration enables developers to export traces, logs, and metrics from…

Mike X Cohen's new book "50 ML Projects to Understand LLMs" uses GPT-2 as a scientific specimen, teaching readers to investigate the model through 50 hands-on projects focused on code, statistics, and controlled experime…

NVIDIA has introduced the NVFP4 training recipe in TransformerEngine, enabling 4-bit mixed-precision pre-training on Blackwell GPUs with no measurable accuracy loss compared to FP8 baselines. The recipe, demonstrated in …