Mixture-of-experts routing

U-Mamba

U-Mamba: Enhancing Long-range Dependency for Biomedical Image Segmentation

Superseded baseline#240 of 1,370 most-superseded · first seen Jan 9, 2024

Cited as a baseline — critiqued by newer work, not yet beaten on a benchmark here

2 papers critique it · 0 beat it on benchmarks

What papers say

Verbatim critique sentences, each from a paper that cites U-Mamba as a baseline.

these models often lack effective cross-scale interaction due to their unidirectional information flow
M2Restore: Mixture-of-Experts-based Mamba-CNN Fusion Framework for All-in-One Image Restoration
State-space alternatives such as U-Mamba and Swin U-Mamba reduce complexity for large images via linear-time sequence modeling, but fine boundary precision can remain challenging in difficult regions.
Shape-Adapting Gated Experts: Dynamic Expert Routing for Colonoscopic Lesion Segmentation

What to use instead

Recent methods in the same sub-problem, not yet superseded in the knowledge base — arXiv benchmark leaders, not vetted production recommendations.