Tenobrus on X: "for some reason people seem to have Anthropic Derangement Syndrome over a pretty straightforward combination of cryptographic primitives and LLM sampling techniques. so as someone who isn't an employee, let me try to explain what's going on with watermarking step by step. -" / X

for some reason people seem to have Anthropic Derangement Syndrome over a pretty straightforward combination of cryptographic primitives and LLM sampling techniques. so as someone who isn't an employee, let me try to explain what's going on with watermarking step by step. -

for some reason people seem to have Anthropic Derangement Syndrome over a pretty straightforward combination of cryptographic primitives and LLM sampling techniques. so as someone who isn't an employee, let me try to explain what's going on with watermarking step by step. - LLMs work by probabilistically autoregressively sampling tokens. meaning: at every token, the model weights don't output a single next token, but a probability distribution over *all possible tokens*, a bunch of little numbers that sum to 1 - the "temperature" sampling setting affects how this distribution is…

saved by

• Tina

related reading

• kat (@transkatgirl) on Xx.com
• How Claude's text watermarking worksanthropic.com
• Defeating Nondeterminism in LLM Inference - Thinking Machines Labthinkingmachines.ai
• Guardian Angels: LLM Personalization for Productivity and Security · Gwern.netgwern.net
• The Waluigi Effect (mega-post) — LessWronglesswrong.com
• GenAI Handbookgenai-handbook.github.io
• How LLMs Work, Explained Without Math - miguelgrinberg.comblog.miguelgrinberg.com
• String Seed of Thought: Prompting LLMs for Distribution-Faithful and Diverse Generationpub.sakana.ai
• LLM Samplers Explained · GitHubgist.github.com
• Transformer Explainer: LLM Transformer Model Visually Explainedpoloclub.github.io
• [2301.10226] A Watermark for Large Language Modelsarxiv.org
• The Mark of the Machinenewyorker.com