CAMELoT: Towards Large Language Models with Training-Free Consolidated Associative Memory

Zexue He, Leonid Karlinsky, Donghyun Kim +3

Large Language Models (LLMs) struggle to handle long input sequences due to high memory and runtime costs. Memory-augmented models have emerged as a promising solution to this prob…