Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
A project called Mini-AGI is training a dynamic continual learning model on 8GB of VRAM by combining dynamic Mixture of Experts, which adds and prunes experts during training, with batch-1 training on a single continuous…
Multiverse's new paper, "LLM Compression by Block Removal with Constrained Binary Optimization" (arXiv 2602.00161), reformulates block-removal pruning of large language models as a constrained binary optimization problem…
Researchers reported in an arXiv paper submitted on 18 Sep 2026 that Apple MacBooks' built-in inertial measurement unit (IMU) can be read without root access via an IOKit driver, enabling a side-channel attack they call …
Decart and Etched's Oasis, a Minecraft-like game powered entirely by a foundation model rather than a physics engine, demonstrates next-frame prediction using a ViT autoencoder and DiT backbone trained on gameplay footag…
MIT researchers used an AI-powered barcode method to identify aging "zombie cells" in tissue, according to MIT News. The technique targets senescent cells, which accumulate with age and are linked to tissue decline.
Massachusetts Institute of Technology researchers developed a noninvasive method to identify senescent "zombie cells" by combining Raman microscopy with single-cell gene expression analysis, publishing the work September…
Towards Data Science published an article titled "Your AI Assistant Wrote the Code. Who Checked the Defaults?" examining five scikit-learn defaults that the publication says deserve closer scrutiny before a model reaches…
Maths educator Grant Sanderson of 3Blue1Brown has launched a series framing intelligence as compression, arguing that distilling fundamental structure from the real world and relating it usefully to other things is what …
TypeSafe AI released Jev, a model that returns only typed answers — choice, score and probability — rather than generated text, and in a 40-ticket support triage test it ran ten times faster than Claude Fable 5.1 and 329…
A developer shared a week of self-directed study covering machine learning fundamentals, backend basics, and web security. The learning notes covered supervised, unsupervised, and reinforcement learning, exploratory data…
Gemini 3.8 Flash identified spider species correctly on 997 of 2,000 photos, a 49.85% accuracy rate that was the highest among nine AI models tested in a benchmark built from research-grade iNaturalist observations cover…
Spectro Cloud published part 1 of a two-part post defining "hybrid AI" as running each workload on the model and infrastructure that make the most sense, with a consistent operating model across enterprise-controlled and…
Apple's M4 chip, first shipped in the 2024 iPad Pro, is the first publicly available device to support ARM's scalable matrix extension (SME), replacing the proprietary AMX instruction set Apple has used in its devices si…
Cellular Intelligence, the Boston-based startup formerly known as Somite AI, added Moderna co-founder Robert "Bob" Langer and Meta chief AI scientist Yann LeCun to its scientific advisory board, along with Jens Nielsen o…
Researchers introduced FRAUDSkill, a method for audio anti-fraud detection that uses structured frozen-weight skill optimization, according to the paper's headline. The approach targets large audio-language models, which…
A developer has introduced YOalphabet and YOconlang, a spatial 4-bit bytecode script and constructed language that maps visual glyph geometry directly to binary hardware registers, eliminating the need for vector embeddi…
A developer explains unsupervised learning, a machine learning approach in which algorithms identify patterns, groups, and relationships in unlabeled data without predefined answers. The writeup covers clustering techniq…
Researchers introduced MoME, a Mixture-of-Memory Embeddings method for context-aware sparse lookup that addresses a limitation in existing memory-embedding approaches, which retrieve via a deterministic mechanism. The wo…
Researchers at the University of Ottawa and Nokia Bell Labs warn that intent-based networking in AI-native 6G designs opens a new threat they call adversarial intent injection, in which attackers push malicious intents t…
A new research paper addresses on-policy distillation (OPD), a method for improving reasoning models by learning the token-level discrepancy between a stronger teacher model and an on-policy student model. The work argue…