Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
A blog post from the Keras team highlights the growing role of deep learning in astroparticle physics, where observatories spanning thousands of square kilometers generate massive datasets that traditional analysis canno…
A final-year project by a data scientist found that the best-performing fraud detection model among six trained was not deployed in production, highlighting the gap between evaluation metrics and production decisions. Th…
A developer's guide emphasizes verifying technical facts before building a RAG application with Next.js, Qdrant, and OpenAI, warning against treating unverified configurations as production facts. The article advises cre…
A new analysis warns that autonomous prior-authorization agents in healthcare can fail when vector databases retrieve expired payer policies, causing denials. The article describes a case where an agent cited a retired 2…
A new multimodal foundation model called oFM, developed on a real-world oncology cohort of 1.67 million cancer patients, achieved a three-fold higher treatment-benefit AUTOC than baseline features across 11 comparative-t…
Pydantic's open-source tool Braindump extracts coding rules from GitHub PR review comments and generates AGENTS.md files, converting 4,668 review comments on pydantic/pydantic-ai between October 2025 and February 2026 in…
Deepgram released Enhanced Metrics and Prometheus/OpenTelemetry support for its speech-to-text and text-to-speech models on Amazon SageMaker AI, enabling customers to monitor billing, feature usage, and per-GPU engine me…
AWS, NVIDIA, and Heidi Health report that using NVIDIA CUDA Multi-Process Service (MPS) with NVIDIA Triton Inference Server on Amazon EC2 GPU instances cuts ASR inference GPU infrastructure requirements by 75%, from 16 i…
A developer built a local knowledge base on Actian VectorAI DB, emphasizing that the database does not embed text but stores vectors and runs nearest-neighbor searches. The project separates concerns: an embedding model …
Wiz, a cloud security company, built an AI-powered data classification system that evolved from a public bucket scanner into a context engine running across all data surfaces, including IaaS and SaaS. The system uses mul…
JetBrains Research has open-sourced DPTrainer, a library that integrates Opacus with Hugging Face's Trainer to enable differential privacy (DP-SGD) training without rewriting training loops or modifying trainer source co…
Artificial intelligence can detect signs of heart disease in women from routine mammograms, according to findings presented at the European Society of Cardiology's annual congress in Munich. Researchers in Israel analyze…
Tether's AI team has developed dedicated, resource-optimized translation models that run entirely on-device, requiring only 21-35 MB per language pair and translating at ~46ms per sentence, which is ~78 times faster than…
Independent testing of the open-source ConceptNet enterprise voice intent classifier, developed by wushu75 and hosted on Hugging Face as conceptnetUk/intent-classifier, reproduced its reported 98.6% accuracy across 9 lan…
Researchers introduced Tensor Mixture (MixT), a general tensor-structured compression scheme that replaces dense linear layers in large language models with natively executable mixtures of tensor operators, and evaluated…
Washington State University researchers used AI to search more than 100 million possible 3D-printing configurations for the NASA alloy GRCop-42, identifying six successful settings after just 40 experiments, including on…
A developer's production checklist for LLMOps in RAG systems identifies weak retrieval, lack of reranking, poor prompt/version control, insufficient eval coverage, and blunt cost controls as common root causes of halluci…
Konjugate's causal inference tool now supports interaction terms and input replay, enabling models to represent relationships where the effect depends on multiple variables jointly, such as predator-prey dynamics or mass…
Researchers proposed FLINT, a workload-driven high-bandwidth flash (HBF) substrate that integrates 3D-stacked NAND flash as a memory-capacity tier alongside HBM to accelerate large language model (LLM) inference, address…
A developer has published a walkthrough demonstrating how to use SynapCores as a unified backend for both LlamaIndex vector and property graph stores, eliminating the need for separate databases. The integration packages…