Wisdom of Committee: Distilling from Foundation Model to Specialized Application Model

TL;DR

This paper introduces a novel knowledge distillation approach using a committee of foundation and complementary teachers, called DiverseDistill, to effectively transfer knowledge to specialized application models, improving their performance.

Contribution

It proposes a teaching committee framework with diverse teachers and a new distillation method, DiverseDistill, to bridge the gap between foundation and specialized models.

Findings

Complementary teachers improve student performance.

DiverseDistill outperforms baseline distillation methods.

Enhanced knowledge transfer across model disparities.

Abstract

Recent advancements in foundation models have yielded impressive performance across a wide range of tasks. Meanwhile, for specific applications, practitioners have been developing specialized application models. To enjoy the benefits of both kinds of models, one natural path is to transfer the knowledge in foundation models into specialized application models, which are generally more efficient for serving. Techniques from knowledge distillation may be applied here, where the application model learns to mimic the foundation model. However, specialized application models and foundation models have substantial gaps in capacity, employing distinct architectures, using different input features from different modalities, and being optimized on different distributions. These differences in model characteristics lead to significant challenges for distillation methods. In this work, we propose creating a teaching committee comprising both foundation model teachers and complementary teachers. Complementary teachers possess model characteristics akin to the student's, aiming to bridge the gap between the foundation model and specialized application models for a smoother knowledge transfer. Further, to accommodate the dissimilarity among the teachers in the committee, we introduce DiverseDistill, which allows the student to understand the expertise of each teacher and extract task knowledge. Our evaluations demonstrate that adding complementary teachers enhances student performance. Finally, DiverseDistill consistently outperforms baseline distillation methods, regardless of the teacher choices, resulting in significantly improved student performance.