Formula

Sequence Extension with a Sampled Token

In autoregressive text generation, a new sequence is formed at each step by appending a newly generated token to the existing sequence. If a token yˉi\bar{y}_i is sampled at step ii, the new sequence, denoted yˉ\bar{\mathbf{y}} (or sometimes ), is constructed by concatenating the preceding sequence y1...yi1y_1...y_{i-1} with this sampled token. This process is formally represented by the equation: yˉ=y1...yi1yˉi\bar{\mathbf{y}} = y_1...y_{i-1}\bar{y}_i

0

1

Updated 2026-05-02

Contributors are:

Who are from:

Tags

Ch.5 Inference - Foundations of Large Language Models

Foundations of Large Language Models

Foundations of Large Language Models Course

Computing Sciences