Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
A deep-dive walkthrough documents ML.NET, Microsoft's open-source, cross-platform machine learning framework that lets .NET developers build, train, and run models entirely in C# or F# without leaving the .NET ecosystem.…
A paper submitted to arXiv on 6 Oct 2026 reports that Station, an open-world environment where multiple agents simulate a scientific ecosystem, rediscovered 62.7% of the findings criteria from three recent ICLR oral pape…
A developer built Touch Grass Bingo, an offline nature bingo web app that uses the open-weight CLIP ViT-B/32 vision model, 8-bit quantized to about 150 MB, running entirely in the phone's browser via Transformers.js on O…
A developer released Whistle, a local speech-to-text model that is 16.9 MB — small enough to bundle inside a desktop installer, browser extension, or Raspberry Pi image — compared with OpenAI's Whisper tiny at roughly 75…
VideoDB's Action Segmentation + Search pipeline returned a relevant moment in its top 10 results for 40.39% of 11,369 text queries on a 100-hour benchmark of 2,835 egocentric videos, about 5 points ahead of AWS Nova (35.…
Researchers Mohammad Hanan Gani and colleagues posted a paper to arXiv on 8 October 2026 introducing "mental-model-enabled agents," a framework that learns an amortized recursive Theory-of-Mind representation with first-…
Training GPT-2 small-style models with low-rank factorised embeddings on the input side alone reduced test loss, while applying the trick to both embeddings and the output head raised loss by 0.07, 0.10 and 0.14 across t…
A developer applied DBSCAN clustering with Hamming distance as the metric to NVIDIA CUDA SASS binaries, extracting instruction words from cubin files via cuobjdump and compiling cuda-samples for sm_50. Working with a dat…
Paris-based AI biology lab Living Models paired Google's Gemma 4 E4B with BOTANIC-1, a plant genomic foundation model pre-trained across 320 plant species, to compress crop genetics research that traditionally takes year…
Liquid AI released two open-weight d1 "decision models" on Hugging Face — d1-3B for text and images and an experimental d1-omni-600M for text plus images or audio — which return typed, calibrated answers (yes/no probabil…
A feasibility study of 150 patients at increased risk of basal cell carcinoma found that AI-assisted line-field confocal optical coherence tomography (LC-OCT) imaging identified histologically confirmed subclinical tumor…
A developer's ALHR (Adaptive Learnable Hierarchical Routing) sparse attention system, which uses static binary trees and learnable functions to reduce keys read per query, achieved 92.1% top-1 accuracy on the MQAR test a…
A developer re-derived results from a manuscript that measures the cost of imposing proportional per-group fairness quotas, finding that on real data every one of 24 cases had a non-empty free region with a median free w…
A developer built FrostWise, an offline frost-date prediction and planting-advice tool that runs entirely on a local machine with no API key, account, or internet connection. The project pairs Prior Labs' TabPFN-v2 tabul…
Nvidia granted Nebius Exemplar Cloud validation for its production HGX B300 racks on October 8, 2026, confirming the systems Nebius rents to customers match Nvidia's reference architecture for training workloads. Nebius,…
A benchmark by Jules Belveze found that a fine-tuned RoBERTa-base model (125M parameters) scored 0.783 macro-F1 on a 22k-item TLDR newsletter classification dataset, beating the best zero-shot decision model, Clef (27B p…
A developer walked through building and training a neural network in code using tf.keras.Sequential, describing an input layer, three Dense hidden layers with ReLU activation, and a 3-neuron softmax output layer. The net…
A Show HN project called awareliquid_physics released a physics-informed machine learning model that writes a learned Hamiltonian advanced by a symplectic integrator directly into its architecture, conserving energy by c…
Scientists at the University of California, San Francisco demonstrated the first single brain implant that decodes a paralyzed person's attempted speech and gestures simultaneously, converting them into text and digital-…
A developer introduced OnTrack, a real-time agent monitoring system that uses streaming optimal transport to compare an agent's unfolding execution trajectory against reference runs, flagging divergence in roughly a mill…