MODE: Multi-Objective Adaptive Coreset Selection

TL;DR

MODE is a dynamic coreset selection framework that adapts strategies throughout training to improve data efficiency, model performance, and interpretability.

Contribution

It introduces a multi-objective adaptive approach to coreset selection that adjusts criteria during training phases, unlike static methods.

Findings

Achieves (1-1/e)-approximation with O(n log n) complexity.

Demonstrates competitive accuracy with reduced memory usage.

Provides interpretable insights into data utility evolution.

Abstract

We present Mode(Multi-Objective adaptive Data Efficiency), a framework that dynamically combines coreset selection strategies based on their evolving contribution to model performance. Unlike static methods, \mode adapts selection criteria to training phases: emphasizing class balance early, diversity during representation learning, and uncertainty at convergence. We show that MODE achieves (1-1/e)-approximation with O(n \log n) complexity and demonstrates competitive accuracy while providing interpretable insights into data utility evolution. Experiments show \mode reduces memory requirements