TL;DR Today we release LeRobotDataset:v3! In our previous LeRobotDataset:v2 release, we stored one episode per file, hitting file-system limitations when scaling datasets to millions of episodes. LeRobotDataset:v3 packs multiple episodes in a single file, using relational metadata to retrieve information at the individual episode level from multi-episode files. The new format also natively supports accessing datasets in streaming mode, allowing to process large datasets on the fly.We provide a one-liner util to convert all datasets in the LeRobotDataset format to […]

We’re thrilled to share that Public AI is now a supported Inference Provider on the Hugging Face Hub! Public AI joins our growing ecosystem, enhancing the breadth and capabilities of serverless inference directly on the Hub’s model pages. Inference Providers are also seamlessly integrated into our client SDKs (for both JS and Python), making it super easy to use a wide variety of models with your preferred providers. This launch makes it easier than ever to access public and sovereign […]

Building trust in the open model ecosystem through standardized risk assessment More than 500,000 models can be found on the Hugging Face hub, but it’s not always clear to users how to choose the best model for them, notably on the security aspects. Developers might find a model that perfectly fits their use case, but have no systematic way to evaluate its    

We’re thrilled to share that Scaleway is now a supported Inference Provider on the Hugging Face Hub! Scaleway joins our growing ecosystem, enhancing the breadth and capabilities of serverless inference directly on the Hub’s model pages. Inference Providers are also seamlessly integrated into our client SDKs (for both JS and Python), making it super easy to use a wide variety of models with your preferred providers. This launch makes it easier than ever to access popular open-weight models like gpt-oss, […]

In an ideal world, AI agents would be reliable assistants. When given a query, they would easily manage ambiguity in instructions, construct step-by-step plans, correctly identify necessary resources, execute those plans without getting sidetracked, and adapt to unexpected events, all while maintaining accuracy and avoiding hallucinations. However, developing agents and testing these behaviors is no small feat: if you have ever tried to debug your own agent, you’ve probably observed how tedious and frustrating this can be. Existing evaluation environments […]

When we think about building a model - be it a Large Language Model (LLM) or a Small Language Model (SLM) - the first thing we need is data. While a vast amount of open data is available, it rarely comes in the exact format required to train or align models. In practice, we often face scenarios where the raw data isn’t enough. We need data that is more structured, domain-specific, complex, or aligned with the task at hand. Let’s look at some common […]

TL;DR: This work shows how a lightweight vision–language model can acquire GUI-grounded skills and evolve into an agentic GUI coder. We release all training recipes, data-processing tools, resulting model, demo and datasets to enable full reproducibility and foster further research 🫡. Find the collection here. This video demonstrates the model obtained through the recipe described below, executing a task end-to-end. Table of Contents

We released swift-transformers two years ago (!) with the goal to support Apple developers and help them integrate local LLMs in their apps. A lot has changed since then (MLX and chat templates did not exist!), and we’ve learned how the community is actually using the library. We want to double down on the use cases that provide most benefits to the community, and lay out the foundations for the future. Spoiler alert: after this release, we’ll focus a lot […]

実世界分布に基づいた日本人ペルソナのための複合AIアプローチ 日本の AI の未来に向けたオープンデータ 高品質で多様なトレーニングデータなしに、日本文化を真に理解するAIを構築することはこれまでほぼ不可能でした。これを変えるため、NVIDIAは、日本の人口統計、地理的分布、文化的特性に沿ったペルソナを含む初のオープン合成データセット、Nemotron-Personas-Japan を公開しました。CC BY 4.0 ライセンスのもと提供される本データセットは、機微な個人データに依存することなく日本社会を反映した AI システム構築のための、プライバシー保護と規制対応を両立した基盤を提供します。 NVIDIA のエンタープライズ向け合成データ生成システム、NeMo Data Designer を用いて作成されたNemotron-Personas-Japan は、すでに広く利用されている US Personas データセットの成功を機に日本版として開発されました。本リリースは、各国・地域におけるソブリン AI 開発を支援する合成ペルソナデータセットとデータ構築方法のグローバルコレクションの第一弾です。 本データセットは、Nemotron モデルをはじめとするオープンソースの 大規模言語モデル(LLM) とシームレスに連携するよう設計されており、企業向けチャットボットから各種ドメインの AI エージェントに至るまで、日本語 AI アプリケーション向けのファインチューンを容易に行えるようになっています。 データセットの内容 合計600万件（各レコードにつき6ペルソナ、100万レコード）の自然な日本語で記述されたペルソナ 1レコードあたり22項目：6つのペルソナ関連項目と、公式の人口統計・労働統計に基づいた16のコンテキスト項目 総トークン数約14億：そのうち約8億5000万がペルソナ関連トークン 約95万件の固有の名前：合成データ生成で前例のない多様性    

People are trying to vibe code games. And it kind of works, at first. However, as the project grows, things begin to fall apart. Why? And what can we do about it? I’ll talk about the problem, how I fixed it, and where to go from here.