4 papers
Beyond Geometric Complementarity: Coherent Overlap in Sparse Mixture-of-Experts Routing
Huiyuan Tian, Bonan Xu, Shijian Li
Sparse mixture-of-experts (MoE) language models route each token to multiple experts, suggesting a geometric account of their benefit: co-selected experts should contribute distinc…
Distillation Dynamics: Towards Understanding Feature-Based Distillation in Vision Transformers
Huiyuan Tian, Bonan Xu, Shijian Li
While feature-based knowledge distillation has proven highly effective for compressing CNNs, these techniques unexpectedly fail when applied to Vision Transformers (ViTs), often pe…
From Per-Image Low-Rank to Encoding Mismatch: Rethinking Feature Distillation in Vision Transformers
Huiyuan Tian, Bonan Xu, Shijian Li
Feature-map knowledge distillation (KD) transfers internal representations well between comparably sized Vision Transformers (ViTs), but it often fails in compression. We revisit t…
SpectralKD: A Unified Framework for Interpreting and Distilling Vision Transformers via Spectral Analysis
Huiyuan Tian, Bonan Xu, Shijian Li +1
Knowledge Distillation (KD) has achieved widespread success in compressing large Vision Transformers (ViTs), but a unified theoretical framework for both ViTs and KD is still lacki…