Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
TileLang introduces a middle-ground approach for writing high-performance GPU kernels, offering explicit control over shared memory allocation, pipeline staging, and warp partitioning through Python code while automating…
SkillOpt introduces a method for improving AI agent performance by treating skills as external state, allowing a frozen agent to be evaluated on scored batches while a separate optimizer model proposes structured edits. …
DeepSeek researchers introduced manifold-constrained hyper-connections to restore the identity mapping property in transformer architectures, addressing training instability and scalability issues caused by standard hype…
Michellebuchiokonicha published Part 1 of "The Machine Learning Engineering Series," detailing the full lifecycle of how raw data becomes a trained model, is containerized, and ships as a live API for other engineers to …
A developer has introduced the AI Development Life Cycle (AIDLC), a structured framework designed to address the unique challenges of machine learning systems that traditional software development life cycles (SDLC) cann…
A developer achieved a 100% join rate on a 96-record Crunchbase-to-CRM dataset by replacing exact string matching with fuzzy matching using the RapidFuzz library, up from a 58% rate with normalized exact matching. The pi…
Cognicore has released NEXUS, an open-source framework that provides a standardized memory, retry logic, and orchestration layer for AI agents, eliminating the need for developers to rebuild these components from scratch…
Researchers have developed Contrastive Decoding Diffing (CDD), a method that recovers verbatim facts from finetuned language models using only output-level logit distributions, without requiring access to model weights o…
Google announced at I/O 2026 that developers can now build AI-powered features directly into web applications using Chrome's built-in AI APIs, eliminating cloud costs and keeping user data on-device. The new capabilities…
Seven open-weight frontier models, including Hugging Face's SmolLM3, DeepSeek-R1, and OpenAI's gpt-oss-120b, were analyzed to distill common training methodologies for multi-billion parameter models, with a focus on arch…
AirSense AI is building a self-improving orchestration layer for its hyper-local air quality intelligence dashboard using Hermes Agent by Nous Research. The agent solves IoT data pipeline challenges by automatically lear…
Researchers have developed ThriftAttention, a mixed-precision attention algorithm that selectively computes only 5% of query-key blocks in FP16 precision while processing the remaining 95% in FP4, recovering 89.1% of the…
Developer Bibhupradhan built FairLens AI, a SaaS platform that automatically audits datasets for hidden biases before they are used to train machine learning models. The tool calculates fairness metrics like Demographic …
Companies building AI applications for industry-specific use cases must choose between retrieval-augmented generation (RAG) and fine-tuning to improve model performance. RAG enables AI models to access real-time informat…
A developer explains that understanding seven core AI metrics—parameters, tokens, FLOPS, TOPS, and FLOPs—is essential for avoiding costly deployment mistakes, such as choosing a 70B-parameter model that runs on a laptop …
Harsh Manvar published a practical guide demonstrating how to run large language models, AI agents, and the Model Context Protocol within Docker containers. The guide provides step-by-step instructions for containerizing…
A new analysis of over 5,400 AI models reveals that benchmark scores for large language models are highly correlated, with just five subjects on the MMLU test predicting the remaining 52 with 91% accuracy. Researchers ha…
A developer has outlined an enterprise decision intelligence architecture that treats binary classification thresholds as operational controls rather than mere technical parameters. The architecture shifts focus from mod…
Ollama released version 0.30.0-rc23, a pre-release that shifts the software's architecture to directly support llama.cpp instead of building on top of GGML, and adds compatibility with the GGUF file format. The update al…
A new open-source tool, Auto GPU Kernel, autonomously discovers and optimizes GPU kernels, achieving a 34.93x average speedup to rank first in the DeepSeek Sparse Attention track of the MLSys 2026 FlashInfer AI Kernel Ge…