liveBook · Manning

Special token

As explained in Effective Conversational AI

A special token is a reserved token with a control or structural meaning in a language-model input or output.

A prompt can contain instructions, examples, and conversation text that blur together, producing poor summaries. Special tokens provide visibly distinct markers for those sections, such as <|instruction|>, <|transcript|>, and <|summary|>. section 12.3.1

Separating a handoff summary prompt

A one-shot summarization prompt places instructions between <|instruction|> markers, the conversation between <|transcript|> markers, and the requested result after <|summary|>. This improved separation produced a focused summary of why the user requested a representative, whereas ordinary colon-based labels produced poor summaries. section 12.3.1

How Effective Conversational AI covers it

How deep it goes
Demonstrates marker placement and output effects, but does not explain tokenization internals or model training.
Examples
Uses one-shot claim-status transcripts marked with instruction, transcript, and summary delimiters.
What it assumes
Assumes familiarity with prompts, transcripts, and language-model summarization while illustrating section separation directly.
What you can do after
Diagnose/debug with it when ambiguous prompt delimiters cause poor summaries or weak separation between instructions and conversation.
How it connects
Connects special tokens with prompt engineering, few-shot learning, transcript summarization, and conversational handoff.
Cover of Effective Conversational AI
liveBookmanning.com
Read Effective Conversational AI on liveBook Get the book at manning.com

Also explained in

Compare how 5 books explain special token →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Effective Conversational AI. The section links point to the parts of the book it draws on.