When Does Resolution Help a Frozen Backbone? Global Attention at Resolution Predicts Scalable Adaptation for Camouflaged and Marine Animal Segmentation

2026-07-02 · Tyler Rust, Chandra Kambhamettu arxiv

Adapting frozen vision foundation models to fine-grained segmentation now largely depends on backbone selection. Whether the backbone applies global attention to a high-resolution token set predicts whether a low-rank adapter turns resolution into accuracy. Isotropic ViTs attend globally over the full grid and keep improving with resolution; hierarchical backbones confine early attention to local windows and pool the grid before their global stages, plateauing at lower resolutions. A controlled six-backbone study establishes the pattern, and editing the backbone points to the cause: pooling keeps the benefit, removing global attention does not. The effect is specific to low-rank adaptation. Under one fixed pipeline, SALT (Side-stem, Attention-gated U-Net, Low-rank Tuning), one RGB-only pass on a strong isotropic backbone wins the best S-measure on the four data-matched camouflaged sets, and leads every marine and salient set. It reaches a new state of the art on both marine-animal benchmarks (MAS3K mIoU 0.878).

📄 PDF Abstract BibTeX arXiv:2607.02708

Code (0)

등록된 구현이 없습니다.

Is Domain Adaptation Always Helpful? A Frozen-Backbone Study of Cross-Domain Sentiment Transfer

2026-07-07 · Phat Tran, Artin Lahni, Pranav Kulkarni, Yaolun Zhang arxiv

Sentiment analysis with frozen pre-trained language model (PLM) backbones has become a common paradigm, yet the practical benefit of explicit domain adaptation remains unclear, particularly when backbones encode varying …

Contrastive Learning Sentiment Analysis Domain Adaptation

DINO-MVR: Multi-View Readout of Frozen DINOv3 for Annotation-Efficient Medical Segmentation

2026-05-08 · Wei Jiang, Feng Liu, Nan Ye, Hongfu Sun arxiv

Adapting foundation models to medical segmentation typically requires either backbone fine-tuning or high-capacity task-specific decoders, both of which are difficult to fit reliably when annotations are scarce. We show …

Tumor Segmentation

When does fusing hand-crafted knowledge with learned representations pay? A cost-normalized benchmark of stacking, substitution, and interference

2026-08-21 · Ahmad AlMughrabi, Albert Clop, Benjamin Busam, Ricardo Marques 외 arxiv

Fusing prior knowledge with data-driven learning is attractive where data is scarce, yet no controlled account says when it helps, is redundant, or harms. We benchmark one fixed hand-crafted knowledge source, a pinned ba…

Encoder Winners Do Not Reliably Transfer Across VLA Backbone Scale: A Frozen-Backbone Grafting Diagnostic

2026-06-12 · Qingping Zeng, Fei She arxiv

Vision-language-action (VLA) policies typically inherit their vision encoder from upstream VLM releases, but it is unclear whether an encoder choice validated on a small VLA transfers to a larger backbone. We introduce a…

Align-RAG: Alignment Is All You Need for TSFM In-Context Learning

2026-08-06 · Mohammad Asadi, Soheil Hor, Bardiya Akhbari, Jack W. O'Sullivan 외 arxiv

Retrieval-augmented forecasting promises to adapt frozen Time Series Foundation Models (TSFMs) to new domains without fine-tuning, but recent methods typically rely on learned fusion modules, i.e., trained adapters that …