CAMELoT: Towards Large Language Models with Training-Free Consolidated Associative Memory
Abstract: Large Language Models (LLMs) struggle to model long input sequences due to high memory and runtime costs. Memory-augmented models have emerged as a promising solution to this problem, but current methods are hindered by limited memory capacity and require costly re-training to integrate with a new LLM. In this work, we introduce an associative memory module which can be coupled to any pre-trained (frozen) attention-based LLM without re-training, enabling effective long language modeling. Unlike previous methods, our associative memory module consolidates representations of individual tokens into a non-parametric distribution model, dynamically managed by properly balancing the novelty and recency of the incoming data. By retrieving information from this consolidated associative memory, the base LLM can achieve significant (up to 29.7\% on Arxiv) perplexity reduction in long-context language modeling compared to other baselines on various standard benchmarks.
This architecture, which
we call CAMELoT (**C**onsolidated **A**ssociative
**M**emory **E**nhanced **Lo**ng **T**ransformer), demonstrates superior performance even with a tiny context window of 128 tokens.
Paper Type: Long
Research Area: Language Modeling
Research Area Keywords: retrieval-augmented models
Contribution Types: Approaches low compute settings-efficiency, Publicly available software and/or pre-trained models
Languages Studied: English
Submission Number: 3542
Loading