18  Faster Decoding

Speculative decoding and its variants, and multi-token prediction.

NoteStatus

Outline. Source: new. See INTEGRATION.md.

18.1 Problem

18.2 Design

18.3 Evolution

18.4 Trade-offs

18.5 Implementation

18.6 Further reading

NoteTODO

Establish the seminal, frontier, and primary-source anchors for this chapter.