Cacheback: Novel Speculative Decoding Method Utilizing CPU Cache

This research explores a novel method for speculative decoding that leverages CPU cache, potentially leading to performance improvements in language models. The paper's novelty lies in its reliance on cache mechanisms, offering a unique perspective on model optimization.

• •Proposes a new speculative decoding technique.
• •Utilizes CPU cache for decoding.
• •Focuses on performance optimization for language models.