Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
Banseok Lee and Youngmin Kim released the official implementation of LittleBit (NeurIPS 2025) and LittleBit-2 (ICML 2026), which compress large language models into the sub-1-bit regime down to 0.1 bits per weight by fac…
Amazon Web Services launched an open-source Physical AI Toolchain for robotics that combines AWS and NVIDIA tools across synthetic data generation, model training, simulation and validation, edge deployment, and continuo…
Towards Data Science published a technical post titled "Your Model's MSE Is Lying to You III: Time Series Diffusion" describing a diffusion head for probabilistic time series forecasting. The post argues that a forecast …
Google DeepMind shipped EmbeddingGemma 2 on October 6, a 740M-parameter open-weight model that encodes text, code, images, audio, and video into a single vector space and runs in 567MB of RAM under an Apache 2.0 license.…
A developer built an offline wake-word detector for a Raspberry Pi 4 voice assistant named "Nova" using openWakeWord and Piper, finding that training positives with Spanish TTS voices instead of English was the single bi…
A developer benchmarked Cloudflare's newly released open-source Clef decision models against TypeSafe's Jev via OpenRouter's decision endpoint, finding Clef tied Jev on accuracy (39/40, 30/30, 12/12) while Clef-flash mad…
Med Karim Bchini ported four jeffhub.ai use cases — ground, support-intents, spam, and triage — onto Google's EmbeddingGemma-2 embedding model, replacing jeffhub's 0.8B "System 1" decider with cosine-similarity scoring o…
A study combining source-resolved magnetoencephalography, spectral parameterization, temporal complexity metrics, and interpretable machine learning found that LSD induces robust, spatially structured increases in alpha …
A developer published a practical guide for ML engineers on building AI-generated image detection pipelines, organizing detection techniques into four families: signal-level methods, metadata and provenance checks, learn…
NVIDIA researchers introduced PivotOPD, an on-policy distillation method that trains multi-turn LLM agents to avoid early pivotal mistakes and recover from them, posting the best average against 13 baselines on 3 agent b…
TM Forum released a portfolio of roughly 20 level-four (L4) autonomy technical solution packages, which it calls the industry's first blueprints for operators to implement self-driving, self-healing networks using AI and…
A developer built a Rubik's Cube Decision Arena that pits six decision models — Laya, Clef, Clef-Flash, GLiNER 2.5, Kev, and Strands Decider — against the identical scrambled cube and one-move question each turn, running…
OpenAI researcher Dan Roberts announced on X that the company's GitHub math repository was updated with 6 new Lean formalizations, 19 modifications, and 3 withdrawals, leaving roughly 42% of top-line results formalized. …
Qdrant released miniCOIL EN-ES, a multilingual extension of its miniCOIL v1 sparse neural retriever, following the v1 article's promise to extend the model to languages beyond English. The company said sparse retrievers …
InclusionAI released Ling 3.1 Flash, a hybrid reasoning mixture-of-experts model with 25B active parameters out of 560B total, now listed on AI Gateway with provider routing, uptime, throughput, latency, activity and Art…
Bitmovin's Video Developer Report, covering ten years of survey data through 2026, shows transcription and speech-to-text ranking first among expected AI/ML use cases every year since 2021 and reaching 48% in 2026, while…
Perplexity released pplx-embed-v2-late in two sizes: a 0.6B model built to run on edge devices and a 9B model for building high-quality indexes, both MIT-licensed and ready to self-host. The models score a best of 92.4% …
MIT associate professor Christina Delimitrou is leading research that applies machine learning to make large-scale cloud computing systems more efficient, aiming to cut data center energy use by improving utilization rat…
A new arXiv paper (2610.09221v1) introduces Consistent Distribution Matching, a simulation-free and data-free distillation method that unifies sample generation and score estimation in one student network alongside a fro…
A new arXiv paper (arXiv:2610.09229v1) introduces Conditional Accuracy Profiling (CAP), a post-hoc diagnostic framework that decomposes pairwise LLM-judge accuracy into eight conditions spanning content sensitivity, robu…