Layer-Specific Prompt Fusion Discovery via Differentiable Search in Vision Foundation Models
Quick Answer
This paper explores layer-specific prompt fusion in Vision Transformers (ViTs) using differentiable architecture search, proposing new fusion methods like affine transformation and cross-attention.
Quick Take
Experiments on 34 datasets demonstrate improved performance over traditional prompt-tuning methods, highlighting the importance of fusion schemes in visual prompt tuning.
Key Points
- Introduces hybrid fusion schemes for visual prompt tuning in ViTs.
- Proposes affine transformation and cross-attention as new fusion methods.
- Achieves consistent performance gains across 34 datasets including VTAB-1k.
- Demonstrates favorable accuracy-latency-parameter trade-off with frozen ViT backbone.
- Highlights the critical role of prompt fusion in enhancing visual prompt tuning.
Paper Resources
📖 Reader Mode
~2 min readAbstract:Visual prompt tuning has emerged as a parameter-efficient fine-tuning approach for adapting large-scale Vision Transformers (ViTs) to downstream tasks. As its learnable prompts are applied in input and feature spaces, prior to jointly going through attention in transformer layers, the most commonly used scheme for fusing image and prompt tokens is concatenation or addition. In this paper, we aim to study a fundamental yet essential problem in visual prompt tuning: whether a single fusion scheme tends to yield better results, and whether that would be beneficial to develop a hybrid fusion scheme. To this end, we formulate the task as a bi-level optimization problem, and solve it leveraging differentiable architecture search. In this context, the learnable prompts and their fusion schemes are jointly optimized. To enrich the search space in the architecture search, we propose two additional fusion schemes, namely, affine transformation and cross-attention, in addition to concatenation and addition. Extensive experiments on 34 datasets spanning VTAB-1k, FGVC, and HTA show consistent gains over prompt-tuning baselines. With a frozen ViT backbone, our method delivers a favorable accuracy--latency--parameter trade-off compared with VPT-Deep and recent variants. Our findings reveal that how prompts fuse with image tokens plays a significant role in visual prompt tuning, and a hybrid fusion fashion can more effectively leverage layer semantics of ViTs, contributing a novel perspective for visual prompt-tuning research.
| Comments: | ECCV 2026 |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV) |
| Cite as: | arXiv:2606.26379 [cs.CV] |
| (or arXiv:2606.26379v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2606.26379 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Xi Xiao [view email]
[v1]
Wed, 24 Jun 2026 21:06:23 UTC (2,335 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CV
See more →ProMoE-FL: Prototype-conditioned Mixture of Experts for Multimodal Federated Learning with Missing Modalities
ProMoE-FL introduces a Prototype-conditioned Mixture-of-Experts framework for multimodal federated learning, effectively addressing missing modalities. It outperforms existing methods on four chest X-ray datasets, demonstrating superior feature synthesis capabilities in both homogeneous and heterogeneous settings.