Depth-Wise Representation Development Under Blockwise Self-Supervised Learning for Video Vision Transformers

TL;DR

This paper explores blockwise self-supervised learning for video vision transformers, demonstrating that it can produce comparable representations to end-to-end training while offering insights into depth-wise representation development and learning dynamics.

Contribution

It introduces a blockwise training method for masked video transformers, analyzing depth-wise representations and comparing it to traditional end-to-end training.

Findings

Blockwise training converges and yields representations close to end-to-end baselines.

Higher-level structure emerges earlier in blockwise training.

Later blocks saturate and maintain geometry, with token-level shifts indicating early mixing.

Abstract

End-to-end backpropagation couples all layers through a global error signal, enabling coordinated learning but requiring long-range credit assignment. Motivated by recent progress in blockwise self-supervised learning (BWSSL), we ask whether masked video transformers can be trained without end-to-end backpropagation. Applying BWSSL to masked video modeling remains relatively underexplored and must handle spatiotemporal context and long-range temporal structure. More broadly, analyses that compare BWSSL and end-to-end training in terms of learning dynamics and depth-wise representation development remain sparse. We apply blockwise learning to a masked autoencoding video vision transformer by partitioning the encoder into blocks, each of which is optimized with a local masked reconstruction loss. Across model sizes and partition granularities, training converges and yields representations close to matched end-to-end baselines under linear-probe and retrieval proxies. In order to compare intermediate representations, we analyze depth-wise decodability, inter-block similarity, and patch-level diagnostics. Blockwise training exposes higher-level structure earlier, while later blocks saturate and operate in a more geometry-preserving regime. It can also induce token-level shifts consistent with stronger early mixing that pooled metrics can miss. These findings point to late-block saturation and interface formation as contributors to the remaining gap.