KV Shifting Attention Enhances Language Modeling

TL;DR

This paper introduces KV shifting attention, a novel mechanism that enhances induction heads in large language models, reducing their complexity and improving language modeling performance and convergence.

Contribution

It proposes KV shifting attention, a new attention mechanism that theoretically reduces the depth and width requirements for induction heads in language models.

Findings

KV shifting attention improves induction head learning.

It leads to better performance or faster convergence in models over 10B parameters.

Theoretical proof supports reduced complexity in induction heads.

Abstract

The current large language models are mainly based on decode-only structure transformers, which have great in-context learning (ICL) capabilities. It is generally believed that the important foundation of its ICL capability is the induction heads mechanism, which requires at least two layers attention. In order to more efficiently implement the ability of the model's induction, we revisit the induction heads mechanism and proposed a KV shifting attention. We theoretically prove that the KV shifting attention reducing the model's requirements for the depth and width of the induction heads mechanism. Our experimental results demonstrate that KV shifting attention is beneficial to learning induction heads and language modeling, which lead to better performance or faster convergence from toy models to the pre-training models with more than 10 B parameters.