UniSpec: Training-Free Speculative Decoding for Robust LLM Acceleration Across Languages and Hardware.

Truong Dinh Do, Nguyen-Khang Le, Le-Minh Nguyen

Browse the full ACL paper archive.