Recasting spatial-spectral coupling as attention scales 3D shape matching to high-res meshes
ssh4net · x · 2026-11-13
A new paper, Spatial-Spectral Coupling as Attention: Scaling Shape Correspondence to High-Resolution Surfaces, reformulates the spatial-spectral coupling at the heart of functional-map methods as scaled dot-product attention: learned descriptors serve as queries and keys, while spectral basis functions act as values.
The key insight is that this coupling—which grows quadratically with mesh resolution—becomes equivalent to attention, so efficient attention implementations can be applied directly to shape matching, providing an effective scaling mechanism. Project page and paper are linked in the post.
More from Research
- One Chart Explains the Linear Algebra Behind Modern AI, from Vectors to SVD — HankYeomans · 2026-11-13
- Codex agent shrinks a 27M-param transformer to 11.46M with equal loss via weight sharing — yoshiyama_akira · 2026-11-13
- Google open-sources Harness²: Agent = Model + Harness, recursive agent harnessing framework — 2C_ornot2C · 2026-11-13
- Question's Gambit: first-step reranking lifts agentic search accuracy to 90.5% — CShorten30 · 2026-11-13
- Terence Tao: major Riemann hypothesis progress came from AI-generated ideas nobody yet understands — burny_tech · 2026-11-13
- Deleting under 0.05% of weights matches full fine-tuning, NYU paper shows — burny_tech · 2026-11-13