Reproduce “ReSS: Residual-Restoring Sparse Attention for 3D Vision Transformers” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Yongsung Kim, Jaehoon Lee, Minjun Park, Wooseok Song, Hun Hwangbo, Sungroh Yoon

ReSS selects attention blocks to remove in multi-view 3D vision transformers by estimating a block’s effect on the residual stream, then refining the dropped set to reduce drift. On π³ at sparsity 0.73, it reaches F1 0.850 on HiRoom, versus 0.378 for SparseVGGT and 0.424 for HeSS, below dense’s 0.908; this is one dataset and operating point.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.