██FR█████ █INTELL███████████
frenchintelligence.org
multi-head-prediction
Unleashing LLM Speed: Multi-Token Self-Speculative Decoding Redefines Inference
July 20, 2025
