SAS: Simulated Attention Score

TL;DR

This paper introduces Simulated Attention Score (SAS), a method that enhances attention mechanisms in Transformers by simulating larger attention capacities without increasing model size, leading to improved performance.

Contribution

The paper proposes SAS, a novel approach to simulate larger attention heads and feature dimensions efficiently, along with PEAA for parameter control, improving Transformer attention performance.

Findings

SAS improves attention performance across multiple datasets.

Simulating larger attention heads enhances model expressiveness.

Parameter-efficient design maintains model size while boosting accuracy.

Abstract

The attention mechanism is a core component of the Transformer architecture. Various methods have been developed to compute attention scores, including multi-head attention (MHA), multi-query attention, group-query attention and so on. We further analyze the MHA and observe that its performance improves as the number of attention heads increases, provided the hidden size per head remains sufficiently large. Therefore, increasing both the head count and hidden size per head with minimal parameter overhead can lead to significant performance gains at a low cost. Motivated by this insight, we introduce Simulated Attention Score (SAS), which maintains a compact model size while simulating a larger number of attention heads and hidden feature dimension per head. This is achieved by projecting a low-dimensional head representation into a higher-dimensional space, effectively increasing attention capacity without increasing parameter count. Beyond the head representations, we further extend the simulation approach to feature dimension of the key and query embeddings, enhancing expressiveness by mimicking the behavior of a larger model while preserving the original model size. To control the parameter cost, we also propose Parameter-Efficient Attention Aggregation (PEAA). Comprehensive experiments on a variety of datasets and tasks demonstrate the effectiveness of the proposed SAS method, achieving significant improvements over different attention variants.