CORTEXA
← Browse
arxivcs.CVcs.AI2026-07-21

Norm or Direction? Decoding Vision Mambas for High-Resolution Vision

Jin Yu, Juyoun Park

Vision Mamba models replace quadratic self-attention with linear complexity selective state space models (SSMs), emerging as efficient visual backbones. However, MambaOut demonstrates that a Gated CNN block can match or exceed VMamba on image classification, questioning the necessity of SSMs for vision. This raises a fundamental question: do VMamba and MambaOut encode visual information differently at the representation level? To investigate, we apply cross model centered kernel alignment (CKA) analysis and find that VMamba's final stage blocks form representations distinctly different from both MambaOut and its own preceding blocks. We therefore focus on the final block features, decomposing each spatial token into magnitude and direction. MambaOut concentrates class-discriminative information in high-norm foreground tokens that align with Grad-CAM attribution. VMamba, by contrast, produces high-norm tokens predominantly in background regions, misaligned with Grad-CAM, yet preserves discriminative signals primarily in token directions. These observations reveal that the two models rely on different encoding strategies. We connect this difference to high-resolution classification and semantic segmentation. VMamba distributes logit support broadly across object regions, whereas MambaOut relies on sparse dominant tokens, a strategy that becomes less stable as token counts grow. Under full fine-tuning for segmentation, VMamba consistently outperforms MambaOut. These results suggest that VMamba's advantage in dense prediction stems not merely from the SSM mechanism or sequence length, but from how semantic evidence is organized across token magnitude, direction. Ultimately, we conclude that token magnitude and directional structure serve as critical axes for improving visual backbones, particularly under dense supervision.

View free PDFSource page

Related papers

arxivcs.CVcs.AI2026-06-26

Mind the Gap: Quantifying the Domain Gap in Cross-Sensor Diffusion Super-Resolution

Dawid Kopeć, Katarzyna Jabłońska, Wojciech Kozłowski, Maciej Zięba

Demand for high-resolution satellite imagery has increased interest in super-resolution (SR) to bridge the spatial resolution gap between freely available missions such as Sentinel-2 and commercial systems like PlanetScope. Because no sensor provides true paired low- and high-res…

View free PDFSource page
arxivcs.CVcs.AI2026-06-28

Fast Enough to Act: Spatio-Temporal Visual Token Merging for Low-Latency Robotic VLMs and VLAs

Junzhou Chen, Jindong Wang, Gang Zhou

Vision-language models and vision-language action models endow the robot with unprecedented capabilities. However, the input of video and high-resolution images yields a massive number of visual tokens, leading to extremely high inference latency and severely hindering the robot'…

View free PDFSource page
arxivcs.CVcs.AI2026-07-15

Human4K: A Large-Scale 4K Multi-View Mocap Dataset for Whole-Body 3D Human Reconstruction

Tianshun Han, Ziyu Shi, Lijian Liu, Ajian Liu, Benjia Zhou, Hugo Jair Escalante, et al.

Recent advances in 3D human reconstruction have improved overall performance, yet current models still fail in the most challenging real-world scenarios. They often produce unstable geometry, inaccurate limb articulation and unreliable predictions under depth ambiguity or self-oc…

View free PDFSource page
arxivcs.CVcs.AIcs.LGcs.MMeess.IV2026-07-21

Mage-Flow: An Efficient Native-Resolution Foundation Model for Image Generation and Editing

Xinjie Zhang, Peng Zhang, Shicheng Zheng, Jinghao Guo, Zhaoyang Jia, Yifei Shen, et al.

Large-scale visual generators are increasingly capable but costly to train, fine-tune, and deploy. We introduce Mage-Flow, a compact 4B-scale generative stack for efficient text-to-image generation and instruction-based image editing. The stack is built from two co-designed compo…

View free PDFSource page
arxivcs.CVcs.AI2026-07-07

EcoVision: AI-Powered Drone Imaging for Salt Marsh Vegetation Monitoring and Dominance Mapping

Innocent Onyenonachi, Peter J. Lawerance, Nadia Kanwal

High-resolution RGB imagery acquired from low-altitude UAV surveys was processed through a modular pipeline incorporating transformer-based semantic segmentation, connected-component vegetation extraction, fine-grained species classification using a ConvNeXt architecture, and gri…

View free PDFSource page
arxivcs.CVcs.AIcs.LG2026-06-26

FedLAS: Feature-Modulated Bidirectional Label Smoothing for Neural Network Calibration

Thiru Thillai Nadarasar Bahavan, Sachith Seneviratne, Saman Halgamuge

Deep Neural Network (DNN) classifiers suffer from poor calibration when their softmax outputs (predictive confidence) deviate from the empirical likelihoods. This manifests itself as either overconfident incorrect predictions or under-confident correct predictions. Label smoothin…

View free PDFSource page