IBM's research demonstrates that large language models alone are insufficient for scalable enterprise AI adoption, requiring "agent logic" — software primitives like knowledge graphs and program analy…
NVIDIA released Cosmos 3, the first open omni-model for physical AI reasoning and action, on Hugging Face June 1, 2026. The single unified model combines world generation, physical reasoning, and acti…
PyTorch released a beginner's guide to its torch.profiler tool, starting with profiling a simple matrix multiplication and addition operation on an A100 GPU. The guide walks through reading profiler t…
Artificial Analysis and IBM Research launched ITBench-AA, the first benchmark for agentic enterprise IT tasks, revealing that all frontier AI models scored below 50% on Site Reliability Engineering ch…
A new free tool called TruthLens, described as a multi-signal deepfake image detector, has been released but its landing page on Hugging Face returns a 404 error. The tool was announced on Hacker News…
Hugging Face researchers released a new method called delta weight sync that reduces the per-step weight transfer in asynchronous reinforcement learning from 1.2 GB to as little as 20 MB for a 0.6B pa…
Pollen Robotics and Hugging Face have released a fully local speech-to-speech pipeline for the Reachy Mini robot, eliminating the need for cloud servers or API keys. The open-source stack runs entirel…
A new glossary from Hugging Face aims to clarify the terms "harness" and "scaffold" in AI agent development, following confusion among researchers at the ICLR 2026 conference. The glossary defines a "…
The article introduces Nemotron-Labs Diffusion, a family of language models that use a diffusion-based approach to generate multiple tokens in parallel and iteratively refine them, offering a faster a…
According to the article, a 3-billion-parameter specialized model outperformed every commercial frontier API tested in a specific enterprise domain at roughly fifty times lower cost, challenging the p…
The article announces the release of OlmoEarth v1.1, a new family of Earth observation AI models that reduces compute costs by up to 3x while maintaining the performance of the original v1 model. This…
Release of OlmoEarth v1.1, a new family of transformer-based AI models designed for processing satellite imagery, which reduces compute costs by up to 3x while maintaining the performance of the origi…
The article announces the release of six new Sentence Transformers CrossEncoder reranker models, ranging from 17 million to 1 billion parameters, which are built on Ettin ModernBERT encoders and achie…
Guide for fine-tuning NVIDIA's Cosmos Predict 2.5 world model using LoRA and DoRA techniques to generate synthetic robot manipulation videos. The approach freezes the base model's 2 billion parameters…
PaddleOCR 3.5 introduces a more flexible inference-engine interface, allowing developers to select the backend (including Transformers) via the `engine` parameter and configure backend-specific option…
The Open Agent Leaderboard is a new open benchmark designed to evaluate the performance and cost of full AI agent systems—including their tools, planning, and error recovery—rather than just the under…
IBM has released two new multilingual embedding models under the Apache 2.0 license, built on ModernBERT: a compact 97M-parameter model and a full-size 311M-parameter model. Both support over 200 lang…
Synchronous continuous batching in LLM inference causes inefficiency by forcing the CPU and GPU to work sequentially, leaving one idle while the other operates. This idle time can account for nearly a…
Convergent infrastructure requirements for the foundation model lifecycle—including tightly coupled accelerator compute, high-bandwidth networking, and distributed storage—and highlights the growing r…
Here is a 2-3 sentence factual summary of the article: The article describes the process of migrating an online reinforcement learning (RL) training system from the vLLM V0 engine to the V1 rewrite, …