Which Direction to Choose? An Analysis on the Representation Power of Self-Supervised ViTs in Downstream Tasks

• URL: http://arxiv.org/abs/2509.15272v1
• Date: Thu, 18 Sep 2025 11:46:07 GMT
• Title: Which Direction to Choose? An Analysis on the Representation Power of Self-Supervised ViTs in Downstream Tasks
• Authors: Yannis Kaltampanidis, Alexandros Doumanoglou, Dimitrios Zarpalas,
• Abstract summary: Self-Supervised Learning for Vision Transformers (ViTs) has recently demonstrated considerable potential as a pre-training strategy for a variety of computer vision tasks.<n>This study aims to bridge the gap by systematically evaluating the use of unmodified features across image classification and segmentation tasks.
• Score: 43.473390101413166
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Self-Supervised Learning (SSL) for Vision Transformers (ViTs) has recently demonstrated considerable potential as a pre-training strategy for a variety of computer vision tasks, including image classification and segmentation, both in standard and few-shot downstream contexts. Two pre-training objectives dominate the landscape of SSL techniques: Contrastive Learning and Masked Image Modeling. Features (or tokens) extracted from the final transformer attention block -- specifically, the keys, queries, and values -- as well as features obtained after the final block's feed-forward layer, have become a common foundation for addressing downstream tasks. However, in many existing approaches, these pre-trained ViT features are further processed through additional transformation layers, often involving lightweight heads or combined with distillation, to achieve superior task performance. Although such methods can improve task outcomes, to the best of our knowledge, a comprehensive analysis of the intrinsic representation capabilities of unaltered ViT features has yet to be conducted. This study aims to bridge this gap by systematically evaluating the use of these unmodified features across image classification and segmentation tasks, in both standard and few-shot contexts. The classification and segmentation rules that we use are either hyperplane based (as in logistic regression) or cosine-similarity based, both of which rely on the presence of interpretable directions in the ViT's latent space. Based on the previous rules and without the use of additional feature transformations, we conduct an analysis across token types, tasks, and pre-trained ViT models. This study provides insights into the optimal choice for token type and decision rule based on the task, context, and the pre-training objective, while reporting detailed findings on two widely-used datasets.

Related papers

• Vision Transformers Need More Than Registers [70.42157905484765]
  Vision Transformers (ViTs) provide general-purpose representations for diverse downstream tasks.<n> artifacts in ViTs are widely observed across different supervision paradigms and downstream tasks.<n>We conclude that these artifacts originate from a lazy aggregation behavior.
  arXiv  Detail & Related papers  (2026-02-25T20:42:35Z)
• Understanding the Transfer Limits of Vision Foundation Models [38.99867932557529]
  Foundation models leverage large-scale pretraining to capture extensive knowledge, demonstrating generalization in a wide range of language tasks.<n>We postulate that this limitation arises from a mismatch between pretraining objectives and the demands of downstream vision-and-imaging tasks.<n>Pretraining strategies like masked image reconstruction or contrastive learning shape representations for tasks such as recovery of generic visual patterns or global semantic structures.<n>Our findings indicate that better alignment between pretraining and downstream tasks, measured by simple divergence metrics such as maximum-mean-discrepancy (MMD) between the same features before and after fine-tuning, correlates with greater performance improvements and
  arXiv  Detail & Related papers  (2026-01-22T12:07:56Z)
• Intra-task Mutual Attention based Vision Transformer for Few-Shot Learning [12.5354658533836]
  Humans possess remarkable ability to accurately classify new, unseen images after being exposed to only a few examples. For artificial neural network models, determining the most relevant features for distinguishing between two images with limited samples presents a challenge. We propose an intra-task mutual attention method for few-shot learning, that involves splitting the support and query samples into patches.
  arXiv  Detail & Related papers  (2024-05-06T02:02:57Z)
• Denoising Vision Transformers [43.03068202384091]
  We propose a two-stage denoising approach, termed Denoising Vision Transformers (DVT) In the first stage, we separate the clean features from those contaminated by positional artifacts by enforcing cross-view feature consistency with neural fields on a per-image basis. In the second stage, we train a lightweight transformer block to predict clean features from raw ViT outputs, leveraging the derived estimates of the clean features as supervision.
  arXiv  Detail & Related papers  (2024-01-05T18:59:52Z)
• KOPPA: Improving Prompt-based Continual Learning with Key-Query Orthogonal Projection and Prototype-based One-Versus-All [24.50129285997307]
  We introduce a novel key-query learning strategy to enhance prompt matching efficiency and address the challenge of shifting features. Our method empowers the model to achieve results surpassing those of current state-of-the-art approaches by a large margin of up to 20%.
  arXiv  Detail & Related papers  (2023-11-26T20:35:19Z)
• Exploring Efficient Few-shot Adaptation for Vision Transformers [70.91692521825405]
  We propose a novel efficient Transformer Tuning (eTT) method that facilitates finetuning ViTs in the Few-shot Learning tasks. Key novelties come from the newly presented Attentive Prefix Tuning (APT) and Domain Residual Adapter (DRA) We conduct extensive experiments to show the efficacy of our model.
  arXiv  Detail & Related papers  (2023-01-06T08:42:05Z)
• Location-Aware Self-Supervised Transformers [74.76585889813207]
  We propose to pretrain networks for semantic segmentation by predicting the relative location of image parts. We control the difficulty of the task by masking a subset of the reference patch features visible to those of the query. Our experiments show that this location-aware pretraining leads to representations that transfer competitively to several challenging semantic segmentation benchmarks.
  arXiv  Detail & Related papers  (2022-12-05T16:24:29Z)
• Self-Promoted Supervision for Few-Shot Transformer [178.52948452353834]
  Self-promoted sUpervisioN (SUN) is a few-shot learning framework for vision transformers (ViTs) SUN pretrains the ViT on the few-shot learning dataset and then uses it to generate individual location-specific supervision for guiding each patch token. Experiments show that SUN using ViTs significantly surpasses other few-shot learning frameworks with ViTs and is the first one that achieves higher performance than those CNN state-of-the-arts.
  arXiv  Detail & Related papers  (2022-03-14T12:53:27Z)
• PreViTS: Contrastive Pretraining with Video Tracking Supervision [53.73237606312024]
  PreViTS is an unsupervised SSL framework for selecting clips containing the same object. PreViTS spatially constrains the frame regions to learn from and trains the model to locate meaningful objects. We train a momentum contrastive (MoCo) encoder on VGG-Sound and Kinetics-400 datasets with PreViTS.
  arXiv  Detail & Related papers  (2021-12-01T19:49:57Z)
• Aligning Pretraining for Detection via Object-Level Contrastive Learning [57.845286545603415]
  Image-level contrastive representation learning has proven to be highly effective as a generic model for transfer learning. We argue that this could be sub-optimal and thus advocate a design principle which encourages alignment between the self-supervised pretext task and the downstream task. Our method, called Selective Object COntrastive learning (SoCo), achieves state-of-the-art results for transfer performance on COCO detection.
  arXiv  Detail & Related papers  (2021-06-04T17:59:52Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.