Rethinking Knowledge in Distillation: An In-context Sample Retrieval Perspective

TL;DR

This paper redefines knowledge in distillation by incorporating relationships between samples and their in-context neighbors, leading to a novel framework that improves performance across various KD paradigms.

Contribution

It introduces a new in-context knowledge distillation framework that leverages sample relationships via retrieval, with theoretical analysis and state-of-the-art results.

Findings

IC-KD outperforms existing methods on CIFAR-100 and ImageNet.

Theoretical analysis confirms the importance of in-context sample knowledge.

Framework is effective across offline, online, and teacher-free KD.

Abstract

Conventional knowledge distillation (KD) approaches are designed for the student model to predict similar output as the teacher model for each sample. Unfortunately, the relationship across samples with same class is often neglected. In this paper, we explore to redefine the knowledge in distillation, capturing the relationship between each sample and its corresponding in-context samples (a group of similar samples with the same or different classes), and perform KD from an in-context sample retrieval perspective. As KD is a type of learned label smoothing regularization (LSR), we first conduct a theoretical analysis showing that the teacher's knowledge from the in-context samples is a crucial contributor to regularize the student training with the corresponding samples. Buttressed by the analysis, we propose a novel in-context knowledge distillation (IC-KD) framework that shows its superiority across diverse KD paradigms (offline, online, and teacher-free KD). Firstly, we construct a feature memory bank from the teacher model and retrieve in-context samples for each corresponding sample through retrieval-based learning. We then introduce Positive In-Context Distillation (PICD) to reduce the discrepancy between a sample from the student and the aggregated in-context samples with the same class from the teacher in the logit space. Moreover, Negative In-Context Distillation (NICD) is introduced to separate a sample from the student and the in-context samples with different classes from the teacher in the logit space. Extensive experiments demonstrate that IC-KD is effective across various types of KD, and consistently achieves state-of-the-art performance on CIFAR-100 and ImageNet datasets.