Software Development

Speculative Decoding

Accelerate LLM inference using speculative decoding, Medusa multiple heads, and lookahead decoding techniques. Use when optimizing inference speed (1.5-3.6× speedup), reducing latency for real-time applications, or deploying models with limited compute. Covers draft models, tree-based attention, Jacobi iteration, parallel token generation, and production deployment strategies.

Publisher / source

Orchestra-Research/AI-Research-SKILLs (found on skills.sh, an awesome list)

Upstream record

Orchestra-Research/AI-Research-SKILLs

Version

Version 3

Published

2026-11-03

Publication review

Approved for publication

Content fingerprint

d818b841ffe3e338569ab7caa0d54516f51c649e52fd1af2d5de5d0435b39f99