Saved in:
Bibliographic Details
Main Authors: Bridge, Oli, Sun, Huey, Branyicskai-Nagy, Botond, D'Ornano, Charles, Basu, Shomit
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2512.03862
Tags: Add Tag
No Tags, Be the first to tag this record!
Table of Contents:
  • Transformer-based architectures have become a dominant paradigm in vision and language, but their success is often attributed to large model capacity and massive training data. In this work, we examine how self-supervised pre-training, intermediate fine-tuning, and downstream fine-tuning interact in a low-capacity regime, using a 5M-parameter Vision Transformer for semantic segmentation. Across multiple data scales, we find that masked image modeling pre-training and downstream fine-tuning reliably improve performance, but with clear diminishing returns as supervision increases. In contrast, inserting an intermediate classification fine-tuning stage consistently degrades downstream performance, with the largest drops occurring precisely where pre-training is most effective. Through an analysis of patch-level representation geometry, we show that classification-based intermediate supervision actively interferes with representations learned during pre-training by collapsing spatial structure critical for dense prediction. These results indicate that, in small models, the geometry of supervision matters more than the number of training stages: misaligned intermediate objectives can negate the benefits of pre-training rather than amplify them.