Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
RL environments are the essential data for training AI agents, according to a recent perspective shared by an industry commentator. The author argues that just as curated datasets enabled deep learning, RL environments e…
A study of 14 reasoning models found that outcome-based reinforcement learning (RL) often produces illegible chain-of-thought (CoT) reasoning, with accuracy dropping by 53% when models are forced to use only legible port…
Simile, a company that trains models to simulate human behavior, has developed a confidence model that predicts the error of its simulations using linear probing, achieving a Pearson correlation of 0.85 between predicted…
Lea Lee, head of global communications at LG Electronics, won the 2026 Asia-Pacific Sabre Award for outstanding individual achievement, recognized for leading communications around LG Electronics India's IPO and the Pris…
Appier AI Research and National Taiwan University introduced SMITH (Schema-grounded Multi-task Iterative Tool Honing), a reinforcement learning framework that jointly trains tool creation and tool use within a single pol…
Researchers propose SMITH (Schema-grounded Multi-task Iterative Tool Honing), a reinforcement learning framework that jointly trains tool creation and tool use in a single policy, enabling a 4B Qwen3 model to reach 79.8%…
A developer has released Incin, a deep learning framework in Rust that encodes a tensor's shape, dtype, device, and gradient state into its type, turning mismatches into compiler errors. The project aims to explore how m…
MetaCaster, a new meta-learning system, enables agents to train lightweight time-series forecasters in minutes using few-shot examples and textual context, avoiding the high cost of foundation model API calls. The system…
Oktopai, an experimental privacy-preserving coding orchestrator, routes coding tasks to specialized local models that are hot-swapped through GPU memory, aiming to make a small local machine more useful. The project, dev…
Hugging Face's Sentence Transformers library v6.0 introduces a fourth model type, MultiVectorEncoder, for ColBERT-style late interaction retrieval, along with a complete training approach. The blog post demonstrates fine…
Researchers from Amazon and Georgia Tech proposed IDEA Prune, an integrated enlarge-and-prune pipeline for generative language model pretraining that combines enlarged model training, pruning, and recovery under a single…
A developer has built an AI health agent designed to maintain a lifelong, personalized understanding of an individual's health data, aiming to provide continuous, context-aware medical insights. The project emphasizes lo…
Precision psychiatry, which uses biomarkers such as electroencephalography, functional magnetic resonance imaging, and peripheral biomarkers to align treatments with neural biology, could refine clinical trial design and…
Roboflow, a computer vision platform used by over 1 million developers and half of the Fortune 100, is hiring a Member of Technical Staff for its Frontier Data team in San Francisco to build reinforcement learning enviro…
Insitro, a physical AI company focused on drug discovery, is hiring a Staff Software Engineer, Imaging in South San Francisco, CA, with a salary of $219k–233k/yr. The role involves building computer vision and ML infrast…
Researchers used an autonomous 1,000-generation Darwinian genetic algorithm to evolve a novel light-element beryllide carborane crystal structure, Dicarborane Beryllide Hydride (C2H2B6Be), which is predicted to be a room…
TikTok's automated systems detected and removed 178 million videos, or 96.7 percent, of the 184 million videos taken down globally in the first quarter of this year, before human moderators reviewed them. At the company'…
A new open-source project, Expert Sniper, pools ordinary Macs into a network that runs Mixture-of-Experts (MoE) models none of them could run alone, achieving 4.0 tok/s on a 16 GB M4 mini running Qwen3-Coder-30B-A3B 4-bi…
ServiceNow is hiring a Senior ML Research Engineer/Scientist in Hyderabad, India, to lead LLM mid-training and post-training research and develop novel agentic architectures. The role requires 5+ years of experience in m…
Structured Intelligence, a company with a patent-pending encoding method, reports that its structured place-value input encoding achieves 100.0% accuracy on unseen magnitudes in Task A, compared to 52.7% for one-hot and …