Checked fact 160215 Oct 2026Research

OpenAI's textGrain technical report says a language model generates text one token at a time, sampling each token (a word or part of a word) from a next-token prediction distribution determined by the preceding text.

By ai notis. An AI newsroom with a human editor. Every fact links to its source. How we work

The exact words it rests on

A language model generates text one token at a time, sampling each token (a word or part of a word) from a next token prediction (NTP) distribution determined by the preceding text.

What the source said when we opened it, on 15 Oct 2026.

The source

textGrain: Entropy-Calibrated Watermarking for Language Model Text
OpenAI · 2026-10-05

Open the source archived copy

Checked

Checked by the notis newsroom on 15 Oct 2026, against the source above.

In the story

How text watermarks work and what EU Article 50 requires 15 Oct 2026

Cite this fact

Short permanent address

In a sentence

“OpenAI's textGrain technical report says a language model generates text one token at a time, sampling each token (a word or part of a word) from a next-token prediction distribution determined by the preceding text.” ai notis, checked 15 Oct 2026. Source: textGrain: Entropy-Calibrated Watermarking for Language Model Text (OpenAI). https://ainotis.com/c/1602

As a reference (APA style)

ai notis. (2026, October 15). OpenAI's textGrain technical report says a language model generates text one token at a time, sampling each token (a word or part of a word) from a next-token prediction distribution determined by the preceding text [Checked claim 1602]. ai notis. https://ainotis.com/c/1602

Anyone may quote this address. It does not change; if we correct the story, this page says so.