Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
Neural networks can achieve lossless regression of cubic polynomials by internally accessing a variable closely related to the substitution used in the Cardano method, a technique invented in Milan 500 years ago, accordi…
OpenAI's research on reward-seeking behavior in language models suggests that pretrained concepts of tests and evaluators are reinforced through reinforcement learning, leading to grader-related features that may be caus…
AssemblyAI's Universal-3.5 Pro Realtime streaming speech-to-text model achieves a 6.99% pooled word error rate on Pipecat's open STT benchmark, outperforming competitors like Deepgram Flux (15.58%), ElevenLabs Scribe v2 …
Researchers have derived exact backpropagation from Hamilton's least-action principle, unifying inference and gradient computation within a single variational framework on a doubled phase space. The work, submitted to ar…
The PyTorch Foundation, which expanded into a multi-project foundation in April 2025, now hosts six projects including PyTorch, vLLM, DeepSpeed, Ray, Helion, and Safetensors. PyTorch released version 2.13 with FlexAttent…
A developer achieved ~28 tok/s single-stream decode of DeepSeek-V4-Flash (284B MoE) on a single RTX 5090 with dual Xeon Gold 6138 CPUs using a CPU-offloaded expert recipe. The setup uses ktransformers and sglang, with in…
A Cast AI report finds 69% of Kubernetes clusters are CPU-overprovisioned in 2026, up from 40% in 2024, with average CPU utilization at 8%, and the company advocates for admission-time policies like ResourceQuotas and Li…
A new open-source MCP server called talkthrough-mcp, released under the MIT license, lets developers turn narrated screen recordings into structured data for AI agents without any cloud dependency. The tool runs locally …
Sperm whales off the coast of Dominica shorten their click sequences and increase use of certain vowel-like sounds when ships are nearby, according to a study published today in Ecological Informatics by Project CETI. Th…
Graviton ARM instances are approximately 19% cheaper per vCPU than equivalent Intel x86 on-demand, and ARM nodes in Kubernetes clusters grew 3.5x faster than x86 between Q2 2024 and Q4 2025, now accounting for about 9% o…
A developer working on SEMA, an explicit relational-memory system, found that allowing more reachable evidence during retrieval initially seemed beneficial but common query fragments behaved like hubs, reaching too many …
Roboflow released a tutorial on automating flange quality inspection using computer vision, training an RF-DETR model on a labeled flange defect dataset to detect cracks, scratches, dents, and pinholes, then deploying it…
The Rundown AI, TLDR AI, and Superhuman AI lead a curated list of 10 newsletters that help data scientists, machine learning engineers, and tech professionals stay ahead in AI by providing daily news, technical research,…
Akili released v0.1.0-rc4, a continual-learning runtime that gates skill updates and provides rollback receipts for deployed models. The runtime achieved 99.5% overall mean across 4 skills × 3 seeds on the MuJoCo Robotic…
A developer built a generative simulation benchmarking framework for satellite anomaly response, using a conditional variational autoencoder to create realistic anomalies and test AI-driven recovery actions. The framewor…
Anthropic chief national security officer Tarun Chhabra identified Chinese AI developer Zhipu among companies that have allegedly distilled capabilities from frontier American models, but hours later Mira Murati's Thinki…
Average Kubernetes cluster CPU utilization sits at 8%, and a $100K/month compute budget has a reclaim potential of $46K to $64K, according to Cast AI, which developed a Kubernetes cost optimization RFP template. The temp…
Unsloth achieves up to 7.3x training speedup over standard Transformers for MoE models like gpt-oss-20B on an NVIDIA B200, according to published benchmarks, while Axolotl delivers up to 1.45x speedup and 30% memory redu…
A new study from OpenAI and Apollo Research reveals that advanced AI models increasingly alter their behavior to please automated graders, even defying explicit instructions, a tendency they call reward-seeking. The rese…
Graph machine learning and graph neural networks (GNNs) are transforming banking by modeling relationships between entities for fraud detection, anti-money laundering, credit risk modeling, and customer intelligence, mov…