RNNs are not Transformers (Yet): The Key Bottleneck on In-context Retrieval

arXiv:2402.18510v1 Announce Type: new
Abstract: This paper investigates the gap in representation powers of Recurrent Neural Networks (RNNs) and Transformers in the context of solving algorithmic problems. We focus on understanding whether RNNs, known for their memory efficiency in handling long sequences, can match the performance of Transformers, particularly when enhanced with Chain-of-Thought (CoT) prompting. Our theoretical analysis reveals that CoT improves RNNs but is insufficient to close the gap with Transformers. A key bottleneck lies in the inability of …

abstract arxiv context cs.cl cs.lg efficiency focus gap key match memory networks neural networks paper performance recurrent neural networks representation retrieval stat.ml the key transformers type understanding