Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
Ferrox v0.23.0, released today, adds support for PrismML's Ternary-Bonsai-2-27B, a 27-billion-parameter model quantized to 1.75 bits per weight that occupies 5.5 GB on disk and runs on a 16 GB laptop, making Ferrox the f…
PrismML's Ternary-Bonsai-2-27B, a 27-billion-parameter reasoning model quantized to 1.75 bits per weight and 5.95 GB on disk, runs on a 16 GB M2 Mac through Frink v0.24.0, a pure-Rust GGUF inference engine that added the…
Figure AI unveiled Helix 2.5 on September 17, 2026, a humanoid control model that achieved a 56% success rate across three household tasks in 30 previously unseen Bay Area homes, completing 237 of 420 attempts against a …
Diogo Almeida, the OpenAI researcher behind RLHF and InstructGPT, released Jev in early access on 15 September through his startup TypeSafe AI. Jev is a non-generative "System One" model that returns typed decisions — a …
Xiaomi's MiMo team began publicly streaming the reinforcement learning training run for its MiMo-V2.6 model, with the MiMo-V2.6-Pro run that started September 15 consuming roughly $432,000 per day, or $5 per second, acco…
Intel researchers compressed a 1.58-bit large language model to an effective 1.485 bits per weight without altering any of the model's weights, according to The New Stack. The technique reduces the model's memory footpri…
An engineer's ablation found that the pi0-FAST vision-language-action model scores 84.7% on LIBERO-10 when a duplicated DeltaActions transform is removed during fine-tuning, versus the 60.5% baseline widely cited from th…
Figure AI introduced Helix 2.5, a neural network pretrained on its Index dataset of human behavior that achieved zero-shot whole-body autonomy across 30 unseen Bay Area homes, performing living-room tidying, towel foldin…
Mattrx replaced its Python scikit-learn microservice with ML.NET running in-process inside its existing .NET 9 application, training regression, classification, and clustering models entirely in C#. The migration cut pre…
A developer argues that explainable AI's core unresolved problem is not computing explanations but communicating them to the people who must act on them, without overstating their certainty. The piece traces the field fr…
Inductive launched Beacon-2 on September 17th, a system that estimates the efficacious human dose of a small molecule directly from its chemical structure by predicting ADMET properties and potency and feeding them into …
AIOps, a term coined by Gartner in 2017, applies AI and machine learning to IT operations to detect anomalies, correlate events, identify root causes, and trigger responses faster than manual processes, according to a Da…
PyTorch 2.14 shipped on September 15 with 2,995 commits from 487 contributors, adding fault-tolerant distributed training as a first-class framework concept via a rewritten NCCL backend (nccl2) with in-place process-grou…
A two-person team built SENTINEL, a single-file, fully offline on-device posture attention map that runs YOLO26n at roughly 16 FPS at 720p on an M4 Mac with WiFi physically disabled. Engineer Igor Eduardo and field speci…
Developer alegrey91 released vex8s, an experimental open-source tool that generates VEX (Vulnerability Exploitability eXchange) documents by correlating container vulnerabilities with Kubernetes securityContext settings …
A developer fine-tuned GLiNER large v2.5 (459M) to 0.83 F1 against Gemini 3.1 Pro's labels for named-entity recognition on Reddit knife comments, after zero-shot GLiNER scored about 0.65 F1. Gemini labeled 4,290 comments…
Developers choosing a model for a small project should prioritize task type and model size over benchmark scores, according to guidance published on Hugging Face's model Hub. The guidance advises picking a model family b…
A tutorial published on Scikit-LLM shows developers how to build a multilingual text classification pipeline using multilingual LLM embeddings and scikit-learn without training a separate model per language. The walkthro…
DataTalks.Club offers five free, community-driven Zoomcamps covering data engineering, machine learning, MLOps, LLMs, and AI development, with course materials available on GitHub for self-paced study. The nine-week Data…
Researchers introduced CERA-MoA, a Mixture-of-Agents framework that co-evolves query routing with continually learning LLM agents, addressing the disconnect in current MoA paradigms where routing and agent fine-tuning ar…