True Positive Weekly #94

The most important artificial intelligence and machine learning news and articles

Hey, in this issue: a revolution in how robots learn; secure LLM tokenizers; extrinsic hallucinations in LLMs; the illustrated AlphaFold; training MoEs at scale with PyTorch; QwQ-32B-Preview; automatic generation of visualizations and infographics; and more.

• [The New Yorker] A revolution in how robots learn

• Secure LLM tokenizers to maintain application integrity

• Extrinsic hallucinations in LLMs

• [NEW BOOK] The Hundred-Page Language Models Book by Andriy Burkov

The first three chapters of Andriy's new book on language models are now online. Over the coming weeks, subsequent chapters will be released. The official release of a top-quality printed edition is planned for January 2025. Subscribe to receive updates on new chapters and on the book's official availability in online stores.

• The illustrated AlphaFold

• Evaluating open LLMs with MixEval: The closest benchmark to Chatbot Arena

• Training MoEs at scale with PyTorch

• [NEW BOOK] Start reading Andriy's new book today. New chapters are coming soon!

• [Model] QwQ-32B-Preview: is an open-weight model with advanced reasoning capabilities

• [Tool] lida: Automatic generation of visualizations and infographics using LLMs

Enjoy the newsletter? Please help us make it bigger and better by sharing it with colleagues and friends.