Research
18 articles
Peer-reviewed papers, technical breakthroughs, and benchmark results from the world's leading AI labs — distilled into plain language. We track arXiv preprints, NeurIPS, ICML, and direct lab publications so you never miss a meaningful advance in AI science.
HuggingFace launches CUGA: lightweight harness for agentic apps
HuggingFace just dropped CUGA — a lightweight harness for building agentic applications, complete with two doz
Google's AMIE Medical AI Matches Doctors in Disease Management
Google's conversational AI system, AMIE (Articulate Medical Intelligence Explorer), just published results in
MolmoMotion: Language-Guided 3D Motion Forecasting Hits HuggingFace
HuggingFace just released MolmoMotion, a model that takes a natural language instruction and predicts a sequen
OpenAI's Deployment Simulation predicts model behavior before launch
OpenAI just rolled out a new technique called Deployment Simulation. It's a method to predict how a language m
DeepMind's DiffusionGemma speeds text generation 4x
DeepMind just dropped DiffusionGemma, a family of text-to-text models that ditch the standard autoregressive a
HuggingFace Details PyTorch Profiling for Fused MLP Layers
HuggingFace just published Part 2 of its PyTorch profiling series, and it's a doozy. The post walks through tr
HuggingFace benchmarks code-switched ASR: OpenAI, Google, Meta fail hard
HuggingFace just released a benchmark specifically for code-switched speech — where speakers mix two languages
Thousand Token Wood: Multi-agent economy runs on a 3B model
Hugging Face released a demo called Thousand Token Wood. It's a simulation where 100 AI agents trade wood, sto
DeepMind's Co-Scientist AI Discovers Genetic Factors That Reverse Cellular Aging
DeepMind's Co-Scientist platform has enabled biologists to identify novel genetic factors capable of rejuvenat
OpenAI's Parameter Golf Challenge Reveals New Frontiers in AI-Assisted Research
OpenAI's Parameter Golf challenge attracted over 1,000 participants who submitted more than 2,000 entries, mak
DeepMind Unveils Decoupled DiLoCo for Fault-Tolerant Distributed AI Training
DeepMind has introduced Decoupled DiLoCo, an advancement in distributed machine learning that enables AI model
DeepMind Unveils AI Co-Clinician to Transform Healthcare Delivery
DeepMind is pioneering research into AI co-clinicians that work alongside healthcare professionals rather than
IBM's Granite 4.1 LLMs: Inside the Architecture of Enterprise AI Models
IBM has released detailed insights into the construction of its Granite 4.1 large language models, designed sp
EMO: New Pretraining Method Unlocks Emergent Modularity in AI Experts
Researchers at HuggingFace have introduced EMO, a novel pretraining approach for mixture of experts (MoE) mode
Google DeepMind Launches Project Genie for AI-Generated Interactive Worlds
Google DeepMind has unveiled Project Genie, an experimental research prototype that enables users to create an
DeepMind Unveils Gemini 3 Deep Think for Advanced Scientific Problem-Solving
DeepMind has announced Gemini 3 Deep Think, an updated version of its most specialized reasoning mode designed
HuggingFace Releases Training Tools for Multimodal Embedding and Reranker Models
HuggingFace has announced new capabilities in its Sentence Transformers library that enable developers to trai
NVIDIA Nemotron 3 Nano Omni: Long-Context Multimodal Model Unveiled
NVIDIA and HuggingFace just released Nemotron 3 Nano Omni, a multimodal model that handles long-context docume