Human-in-the-Loop Mixup

TL;DR

This paper investigates whether synthetic labels used in mixup data augmentation align with human perception, revealing misalignments and proposing a human-in-the-loop approach to improve model robustness and reliability.

Contribution

It introduces the HILL MixE Suite, a set of elicitation interfaces for collecting human judgments on mixup examples, and provides insights into aligning synthetic data with human perception.

Findings

Human perceptions often do not match traditional synthetic labels.

Incorporating human uncertainty can enhance model reliability.

The H-Mix data hub facilitates further research on human-aligned synthetic data.

Abstract

Aligning model representations to humans has been found to improve robustness and generalization. However, such methods often focus on standard observational data. Synthetic data is proliferating and powering many advances in machine learning; yet, it is not always clear whether synthetic labels are perceptually aligned to humans -- rendering it likely model representations are not human aligned. We focus on the synthetic data used in mixup: a powerful regularizer shown to improve model robustness, generalization, and calibration. We design a comprehensive series of elicitation interfaces, which we release as HILL MixE Suite, and recruit 159 participants to provide perceptual judgments along with their uncertainties, over mixup examples. We find that human perceptions do not consistently align with the labels traditionally used for synthetic points, and begin to demonstrate the applicability of these findings to potentially increase the reliability of downstream models, particularly when incorporating human uncertainty. We release all elicited judgments in a new data hub we call H-Mix.