Memory Caching: RNNs with Growing Memory

TL;DR

This paper introduces Memory Caching (MC), a technique that enhances recurrent neural networks by allowing their memory capacity to grow with sequence length, bridging the performance gap with Transformers in recall tasks.

Contribution

The paper proposes Memory Caching (MC), a novel method that enables RNNs to have scalable memory, improving their performance on long-context tasks and reducing the gap with Transformers.

Findings

MC improves RNN performance on language modeling tasks.

MC variants achieve competitive results in recall tasks.

Memory capacity growth enhances RNNs' ability to handle long sequences.

Abstract

Transformers have been established as the de-facto backbones for most recent advances in sequence modeling, mainly due to their growing memory capacity that scales with the context length. While plausible for retrieval tasks, it causes quadratic complexity and so has motivated recent studies to explore viable subquadratic recurrent alternatives. Despite showing promising preliminary results in diverse domains, such recurrent architectures underperform Transformers in recall-intensive tasks, often attributed to their fixed-size memory. In this paper, we introduce Memory Caching (MC), a simple yet effective technique that enhances recurrent models by caching checkpoints of their memory states (a.k.a. hidden states). Memory Caching allows the effective memory capacity of RNNs to grow with sequence length, offering a flexible trade-off that interpolates between the fixed memory (i.e., $O(L)$ complexity) of RNNs and the growing memory (i.e., $O(L^2)$ complexity) of Transformers. We propose four variants of MC, including gated aggregation and sparse selective mechanisms, and discuss their implications on both linear and deep memory modules. Our experimental results on language modeling, and long-context understanding tasks show that MC enhances the performance of recurrent models, supporting its effectiveness. The results of in-context recall tasks indicate that while Transformers achieve the best accuracy, our MC variants show competitive performance, close the gap with Transformers, and performs better than state-of-the-art recurrent models.