text

Definition

Written natural language as a data modality. It appears constantly in compounds that specify input and output formats — text-to-image, text-to-speech, text-to-SQL — and in "plain text" as the interface most tools speak. Practically, raw text never reaches a model directly: it is tokenized first, and tokenization choices affect what the model can represent about spelling, numbers, and rare words.

Explainers using this term

• Paper Walkthrough: The Design Fundamentals of Pixel Text Representation LearningOn the Design Fundamentals of Pixel Text Representation Learning
• Paper Explained — LatentPress: Feeding Compressed Context Straight to a Frozen LLM, Neither as Text Nor as PixelsLatentPress: Context Compression Beyond Text and Vision
• Building a Pretraining Corpus — From Web Sludge to Textbook QualityThe Pile: An 800GB Dataset of Diverse Text for Language Modeling