Break the Sequential Dependency of LLM Inference Using Lookahead Decoding.

Yichao Fu, Peter Bailis, Ion Stoica, Hao Zhang

Browse the full ICML paper archive.