Build a Robot Pick-and-Place Simulation with SimReady Assets
AI summary A walkthrough demonstrates a beverage can pick-and-place task using SimReady assets in Isaac Sim.
News & blogs · AI & ML
Hugging Face Blog
The latest from Hugging Face Blog, part of our AI & ML coverage. Updated automatically as new posts are published.
XYBench: Helping with the problem behind the question
AI summary XYBench, a benchmark of 8,115 questions, tests AI assistants' ability to handle user misconceptions.
llama.cpp v1.0.0 : A stronger foundation for the local AI ecosystem
AI summary Llama.cpp v1.0.0 is released, offering a lightweight platform for local AI with stable APIs and improved infrastructure.
LLMON: An LLM-Native Markup Language, now in Granite 4.1 3B and 8B
AI summary Hugging Face introduces LLMON, an LLM-native markup language to improve structure and semantics in model interactions.
The Definitive Fine-Tuning Guide for 3D Generation Models
AI summary Hugging Face provides scripts and containers to fine-tune TRELLIS.2, a 3D generation model, for specific tasks using the GSO dataset.
BotPsy: On-Device Personal Assistant with a Sub-Billion-Parameter Agentic Language Model
AI summary BotPsy is a 0.8B-parameter model designed for on-device tasks, offering efficient and private personal assistant capabilities.
Token Pooling-Aware Training for Multi-Vector Embedding Models
AI summary Token pooling-aware training helps preserve information when compressing multi-vector embeddings, improving retrieval efficiency.
Pay for Hugging Face Hub with your Google Cloud Account
AI summary Hugging Face now allows organizations to link to Google Cloud accounts for billing purposes.
Decision Models in Hugging Face
AI summary Hugging Face discusses decision models that evaluate options and return probability distributions instead of generating text.
Nemotron-Personas-Finance-Japan: A Synthetic Dataset Capturing Japan's Demographics and Financial Context
AI summary Nemotron-Personas-Finance-Japan is a synthetic dataset combining Japan's demographics and financial data for AI development.
Your Experts Have Partners. There's HOPE for Keeping Them Together
AI summary A new method called HOPE improves expert pruning in MoE models by considering expert interactions.
ToyBlocks: A controlled world for robot learning and evaluation
A Blog post by Ai2 on Hugging Face
Egyptian and Nigerian English join the Open ASR Leaderboard
AI summary Hugging Face adds Egyptian and Nigerian English to the Open ASR Leaderboard through a collaboration with Liva AI.
New NVIDIA Cosmos 3 Open Models for Interactive World Simulation
AI summary NVIDIA released four Cosmos 3 models with 16B parameters for interactive world simulation, supporting robotics, autonomous driving, and research.
A Closer Look at Phonon TTS. Launching High-Quality On-Device Speech That Tops the Benchmarks.
AI summary Phonon is a small on-device text-to-speech model that runs without a network and offers high-quality voice output. It supports multiple languages and custom voices.
Dharma Field Extractor: Any Document In, Structured Fields in JSON Out
AI summary Dharma Field Extractor is a 2-billion-parameter model that extracts specific fields from documents and outputs structured JSON, outperforming other models at lower cost.
StateBench: Testing how language models track state changes over long documents
AI summary StateBench is a benchmark for testing how well language models track state changes in long documents.
Palette: teaching a small model to build decks in your design language
AI summary Palette is a small model trained to create slide decks in a specific design language, outperforming frontier models in some areas.
Transformers Model Integration Contracts
AI summary Hugging Face introduced small model fixtures for faster integration testing, catching bugs like UMT5's attention issue before they reach main.
CWE-bench v2: More coverage, quality, and transparency
AI summary CWE-bench v2 introduces 100 tasks to test cyber capabilities, covering 91 CWEs with programmatic scoring and a held-out set for progress tracking.
nemotron-halloween
AI summary MonsterHouse is an interactive Halloween projection with a grumpy pirate voice, built using NVIDIA tech and a custom animation rig.
Granite Switch 4.1.1 — Update models like you update software
AI summary Granite Switch 4.1.1 is a patch update with incremental improvements and backward compatibility.
Introducing AnySpeechRecognizer and AnyDecisionModel: Open Models for Speech and Decisions on Apple Platforms
AI summary Hugging Face introduces AnySpeechRecognizer and AnyDecisionModel for Apple platforms, offering open models for speech and decisions.
Announcing Project ProtoStar: Making Publicly Funded Science AI-Ready
AI summary Project ProtoStar aims to make publicly funded science more accessible for AI research by promoting open sharing of datasets, models, and code.
Welcome Le Chonk: Mistral Large 4 on Hugging Face
AI summary Mistral Large 4, a multimodal MoE model, is now available on Hugging Face with 1T parameters and 1M token context.
NVIDIA Nemotron 3.5 Transcribe and Granary v2: Intelligent Multilingual Transcription You Can Prompt and Adapt
AI summary NVIDIA Nemotron 3.5 Transcribe is an open-weight model that transcribes speech into text with multilingual support, speaker attribution, and translation.
Build specialized multimodal models with LFM3
AI summary LFM3 is a new generation of Liquid Foundation Models designed for specialized tasks with multimodal support and efficient on-device performance.
Can we trust LLMs to judge whether a research idea is novel?
AI summary A study questions the reliability of LLMs in judging the novelty of research ideas, showing that small changes in evaluation setups lead to different judgments.
LitPivot: Developing research ideas through dynamic contextualization within the literature landscape
AI summary LitPivot helps researchers adjust their ideas based on existing literature, improving idea grounding and novelty.
Does Your Model Run on AWS Trainium? Look It Up
AI summary Hugging Face's TorchNeuron allows Transformers models to run on AWS Trainium, with a Space showing which models work and how to run them.