liveBook · Manning

Value

A value is data or a representation associated with a key or selection context and returned, stored, or aggregated as content.

How 2 Manning books explain it

Rearchitecting LLMs

A value is a vector containing the semantic information that attention aggregates from tokens selected by query-key compatibility.

How it is used
Contributes token information to an enriched embedding according to query–key compatibility.
How deep it goes
Defines the value’s semantic role and explains its weighting through query–key dot products and softmax scores.
Examples
Uses “spring” seeking context from “mountain” in a sentence about fresh water.
What you can do after
Understand how attention weights select value information to enrich a token embedding.

Read how Rearchitecting LLMs explains value →

Machines that Think

A value is an attention representation carrying the content that a token can contribute to another token's representation.

How it is used
Supplies the information tokens can contribute to one another in self-attention and cross-attention.
How deep it goes
Defines values as learned projections and explains their combination into hidden states, without detailing projection mathematics.
Examples
Uses “The sun rises in the east” to illustrate token relevance and encoder information retrieved during prediction.
What you can do after
Understand how values carry information into context-enriched token representations.

Read how Machines that Think explains value →

Also covered in

What to know first

Related concepts

Questions about value

What is value?

A value is data or a representation associated with a key or selection context and returned, stored, or aggregated as content.

Which Manning books explain value?

Rearchitecting LLMs and Machines that Think.

What should I know before learning value?

Manning books that teach value usually introduce array first.

What is value part of?

Value is part of qkv mechanism. The QKV mechanism computes query, key, and value vectors so attention can score relationships and combine contextual information.

These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.