Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
Researchers propose "inoculation pretraining," a method to prevent emergent misalignment in AI systems by adding synthetic training data about good-but-reward-hacking AI personas. The approach aims to increase the prior …
A developer has created a new non-LLM model with strong reasoning capabilities that costs under $10 per month to run, but requires training and guidance like a pet. The creator initially launched the model as part of a g…
Google launched Coral, a full-stack platform for edge AI that combines an AI-first hardware architecture with a unified developer experience to enable efficient, local AI processing. The platform provides software develo…
Hugging Face released Transformers.js, a JavaScript library that runs state-of-the-art NLP models directly in the browser on the user's device with no server required. The library, which is functionally equivalent to Hug…
Researchers have developed Cassandra, an algorithm-hardware co-designed self-speculative decoding framework that enables lossless acceleration of reasoning large language models on edge devices without additional trainin…
An eBPF agent that attaches to the CUDA runtime, CUDA driver, and Linux kernel scheduler simultaneously can trace a GPU stall back to the exact Python source line that triggered it. The tool correlates events across all …
NVIDIA's Blackwell consumer GPU (GB203/SM_120) features a unified TensorCore pipeline where all 12 non-FP64 precision formats share identical 29-cycle latency and 23-cycle throughput, reducing precision selection to a pu…
Researchers have introduced a fixed-budget, cluster-aware standard for evaluating LLM-as-a-Judge systems, using a multi-hop retrieval-augmented generation (RAG) stress test to measure fairness. The new method aims to add…
A new benchmark called BenchBench tests AI models on their ability to create benchmarks for other models, revealing that only GPT 5.2 succeeded in generating a practically solvable yet challenging evaluation. Other leadi…
Model distillation is a standard deep learning technique that trains a smaller "student" model to imitate a larger "teacher" model by matching the teacher's full probability distribution, not just the correct answer. Con…
Cedana, a Y Combinator-backed startup from the S23 batch, is hiring a Forward Deployed Engineer to help maximize AI and HPC cluster utilization through automated GPU checkpointing and live migration. The company’s kernel…
British autonomous-vehicle software startup Wayve is launching a new research unit called Wayve Labs, led by chief scientist Jamie Shotton, to explore embodied intelligence systems beyond self-driving cars. The lab will …
Mallika Rao, a former engineering leader at Twitter, Walmart, and Netflix, warned that "evaluation debt" — the gap between evolving AI system architectures and stagnant testing frameworks — poses a greater risk to produc…
A developer released textsnap, a command-line tool that performs OCR on images, screenshots, PDFs, and webpages entirely on a local CPU without requiring a GPU or cloud connection. The tool uses a quantized 0.9B vision-l…
A developer has built a complete AI agent framework using Anthropic's Claude API that implements the ReAct (Reason + Act) pattern, enabling LLMs to autonomously pursue multi-step goals rather than simply responding to in…
A new video analysis compares spectrograms and MFCCs for audio machine learning, highlighting practical tradeoffs in feature extraction for tasks like speech recognition and music classification. The breakdown examines h…
A software developer built a fully automated document management system using Paperless-NGX that processes over 900 documents for a family of four across multiple languages. The system ingests physical papers via a ScanS…
Researchers at the University of Cambridge and Geodesic Research have proposed a new framework called Developmental Cognitive Interpretability (DCI) that models how an AI system's motivations, goals, and intentions chang…
In 2026, the most effective AI tools for conversion rate optimization are moving beyond traditional A/B testing to build continuous conversion systems that run experiments, personalize experiences, and respond to behavio…
A developer rescued a Retrieval-Augmented Generation (RAG) assistant from memory leaks and deployed it on a Render free-tier instance with a 512MB RAM limit. The engineer re-engineered the pipeline by replacing heavy PyT…