The paper introduces Transformer-Within-Transformer (TWT), a post‑hoc technique that merges contiguous redundant layers in Vision Transformers into a single surrogate layer. By doing so, TWT cuts both parameter count and inference compute while maintaining competitive performance on natural image tasks with only half the depth. In histopathology applications, TWT not only matches but sometimes surpasses the baseline model’s performance.
arXiv:2609.08084v1 Announce Type: cross Abstract: Monocular depth estimation is a ubiquitous yet highly ill-posed computer vision task, with downstream applications in scene reconstruction, computati...
arXiv:2609.37605v1 Announce Type: cross Abstract: Supervised deep learning has advanced sparse-view tomographic reconstruction. However, conventional models, which typically map filtered back-project...
arXiv:2609.23733v1 Announce Type: new Abstract: Feed-forward visual geometry models such as the Visual Geometry Grounded Transformer (VGGT) have recently enabled direct 3D reconstruction from multi-v...
arXiv:2608. 20134v1 Announce Type: cross Abstract: We present a novel view on feature evolution in Vision Transformers (ViTs) by visualizing the training process over two dimensions -- network depth (layer) and training time (epochs).
DART is a new RGB‑D pretraining method for surgical vision foundation models that incorporates pseudo‑labeled depth maps as a pixel‑space reconstruction target during training. By adding a depth reconstruction head to DINOv2’s masked iBOT framework, DART improves representation quality without affecting downstream RGB‑only fine‑tuning or inference. Across eight surgical benchmarks—including segmentation, depth estimation, and image‑level recognition—DART outperforms both natural‑image and in‑domain baselines, demonstrating that geometric pseudo‑labels can strengthen foundation model pretraining without extra labels or inference cost.
arXiv:2609.37631v1 Announce Type: new Abstract: Transformers are typically trained from random initialization, requiring all their capabilities to emerge from large-scale optimization. Recent work sh...
HiLRP introduces a unified attribution framework for Vision Transformers (ViTs) that addresses the challenges posed by diverse architectural designs. By decomposing ViT operations into four basic types—linear maps, bilinear mixing, normalization/gating, and reindexing—HiLRP applies conservation‑satisfying relevance rules, enabling reliable explanations across a wide range of backbones. The method outperforms 14 existing attribution techniques on 10 architectures, maintaining conservation and improving localization accuracy (0.97 Pointing) compared to competitors.
ProgResViT is an input‑adaptive Vision Transformer that processes images progressively across multiple rounds, starting with a low‑resolution image and a narrow subnetwork and refining the prediction with higher resolution and a wider subnetwork if needed. The method introduces Progress‑Conditioned Soft Gating (PSG) to share a single backbone across rounds while conditioning token fusion and layer outputs on the current round, block, and input resolution. Experiments on DeiT show improved accuracy‑compute trade‑offs compared to adaptive‑width, adaptive‑depth, and dynamic‑token baselines, and the design also benefits self‑supervised DINO representations and downstream semantic segmentation.
arXiv:2607. 10391v1 Announce Type: cross Abstract: Despite exposing rich intermediate representations, Vision Transformers (ViTs) are almost exclusively utilized as black-box feature extractors, where only the last layer is considered for downstream tasks.
arXiv:2607. 02612v1 Announce Type: cross Abstract: Vision Transformers achieve strong image classification accuracy but process all image regions with nearly the same computation, even when many regions are redundant or uninformative.
arXiv:2607. 00774v1 Announce Type: cross Abstract: Recent recursive Transformer studies have primarily reused shared parameters across computation steps to construct compact, parameter-efficient models.
arXiv:2603. 18846v3 Announce Type: replace-cross Abstract: Foundation models are used to extract transferable representations from large amounts of unlabeled data, typically via self-supervised learning (SSL).