Investigating Neuron Ablation in Attention Heads: The Case for Peak Activation Centering

TL;DR

This paper explores how different neuron ablation techniques affect transformer models' performance, introducing a novel 'peak ablation' method and analyzing its effectiveness in understanding attention mechanisms.

Contribution

It introduces 'peak ablation' as a new neuron ablation technique and compares it with existing methods to better interpret attention mechanisms in transformers.

Findings

Peak ablation often causes less performance degradation.

Resampling generally leads to more performance loss.

Different ablation methods vary in effectiveness across models.

Abstract

The use of transformer-based models is growing rapidly throughout society. With this growth, it is important to understand how they work, and in particular, how the attention mechanisms represent concepts. Though there are many interpretability methods, many look at models through their neuronal activations, which are poorly understood. We describe different lenses through which to view neuron activations, and investigate the effectiveness in language models and vision transformers through various methods of neural ablation: zero ablation, mean ablation, activation resampling, and a novel approach we term 'peak ablation'. Through experimental analysis, we find that in different regimes and models, each method can offer the lowest degradation of model performance compared to other methods, with resampling usually causing the most significant performance deterioration. We make our code available at https://github.com/nickypro/investigating-ablation.