Attention in Reasoning: Dataset, Analysis, and Modeling

TL;DR

This paper introduces the AiR framework that evaluates and enhances attention mechanisms in neural networks by analyzing reasoning processes, using human data, and supervising attention learning to improve visual question answering performance.

Contribution

It proposes a novel evaluation metric for attention based on reasoning steps, and introduces supervision techniques to improve attention and reasoning in VQA models.

Findings

Attention with reasoning improves task performance.

Supervised attention learning enhances reasoning capability.

Analysis of human and machine attention mechanisms reveals key differences.

Abstract

While attention has been an increasingly popular component in deep neural networks to both interpret and boost the performance of models, little work has examined how attention progresses to accomplish a task and whether it is reasonable. In this work, we propose an Attention with Reasoning capability (AiR) framework that uses attention to understand and improve the process leading to task outcomes. We first define an evaluation metric based on a sequence of atomic reasoning operations, enabling a quantitative measurement of attention that considers the reasoning process. We then collect human eye-tracking and answer correctness data, and analyze various machine and human attention mechanisms on their reasoning capability and how they impact task performance. To improve the attention and reasoning ability of visual question answering models, we propose to supervise the learning of attention progressively along the reasoning process and to differentiate the correct and incorrect attention patterns. We demonstrate the effectiveness of the proposed framework in analyzing and modeling attention with better reasoning capability and task performance. The code and data are available at https://github.com/szzexpoi/AiR