Distilled Circuits: A Mechanistic Study of Internal Restructuring in Knowledge Distillation

2025-05-16

AI Summary

Knowledge distillation’s impact on internal computational mechanisms remains poorly understood, particularly regarding how student models restructure, compress, or discard teacher components. Method: Using GPT2-small and DistilGPT2, we introduce an influence-weighted component alignment metric to quantify functional module alignment post-distillation. We integrate mechanistic interpretability, circuit analysis, activation tracing, and influence functions to assess alignment across multiple tasks. Contribution/Results: We find that distilled students rely on fewer—but more influential—components, challenging the “black-box equivalence” assumption. Although output behavior remains similar, internal computation undergoes significant shifts, degrading robustness and generalization. Our framework provides the first interpretable, quantitative diagnostic tool for assessing functional fidelity in model compression, advancing trustworthy and explainable knowledge distillation.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models
• Data Mining & Knowledge Management: Data Compression

Application Category

• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics
• Web Mining and Content Analysis: Large pretrained models with web data
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Knowledge distillation compresses a larger neural model (teacher) into smaller, faster student models by training them to match teacher outputs. However, the internal computational transformations that occur during this process remain poorly understood. We apply techniques from mechanistic interpretability to analyze how internal circuits, representations, and activation patterns differ between teacher and student. Focusing on GPT2-small and its distilled counterpart DistilGPT2, we find that student models reorganize, compress, and discard teacher components, often resulting in stronger reliance on fewer individual components. To quantify functional alignment beyond output similarity, we introduce an alignment metric based on influence-weighted component similarity, validated across multiple tasks. Our findings reveal that while knowledge distillation preserves broad functional behaviors, it also causes significant shifts in internal computation, with important implications for the robustness and generalization capacity of distilled models.

Problem

• Understanding internal computational transformations in knowledge distillation
• Comparing teacher and student model circuits and representations
• Quantifying functional alignment beyond output similarity in distilled models

Innovation

• Analyze internal circuits using mechanistic interpretability
• Introduce influence-weighted component similarity metric
• Reveal reorganization and compression in student models