Attention Consistency for LLMs Explanation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Lan, Tian, Xu, Jinyuan, He, Xue, Hwang, Jenq-Neng, Li, Lei
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908588699549696
author Lan, Tian
Xu, Jinyuan
He, Xue
Hwang, Jenq-Neng
Li, Lei
author_facet Lan, Tian
Xu, Jinyuan
He, Xue
Hwang, Jenq-Neng
Li, Lei
contents Understanding the decision-making processes of large language models (LLMs) is essential for their trustworthy development and deployment. However, current interpretability methods often face challenges such as low resolution and high computational cost. To address these limitations, we propose the \textbf{Multi-Layer Attention Consistency Score (MACS)}, a novel, lightweight, and easily deployable heuristic for estimating the importance of input tokens in decoder-based models. MACS measures contributions of input tokens based on the consistency of maximal attention. Empirical evaluations demonstrate that MACS achieves a favorable trade-off between interpretability quality and computational efficiency, showing faithfulness comparable to complex techniques with a 22\% decrease in VRAM usage and 30\% reduction in latency.
format Preprint
id arxiv_https___arxiv_org_abs_2509_17178
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Attention Consistency for LLMs Explanation
Lan, Tian
Xu, Jinyuan
He, Xue
Hwang, Jenq-Neng
Li, Lei
Computation and Language
Understanding the decision-making processes of large language models (LLMs) is essential for their trustworthy development and deployment. However, current interpretability methods often face challenges such as low resolution and high computational cost. To address these limitations, we propose the \textbf{Multi-Layer Attention Consistency Score (MACS)}, a novel, lightweight, and easily deployable heuristic for estimating the importance of input tokens in decoder-based models. MACS measures contributions of input tokens based on the consistency of maximal attention. Empirical evaluations demonstrate that MACS achieves a favorable trade-off between interpretability quality and computational efficiency, showing faithfulness comparable to complex techniques with a 22\% decrease in VRAM usage and 30\% reduction in latency.
title Attention Consistency for LLMs Explanation
topic Computation and Language
url https://arxiv.org/abs/2509.17178