Higher cluster tilting objects in locally finite triangulated categories

arXiv:2607.19916 2026 Architecture 1 ideas extracted · analyzed Aug 30, 2026

What the math gives to ML

The paper provides a constructive combinatorial model for higher cluster-tilting objects: indecomposable categorical objects are represented by tagged arcs, morphisms follow local d-moves, and d-orthogonality is detected by geometric non-crossing. This gives a principled way to construct sparse, multiscale interaction graphs rather than selecting attention edges heuristically. The most promising transfer is to use maximal compatible arc families as sparse attention masks and rotate or mutate these families across layers so that each layer is O(N) while the composition gains global receptive field. The categorical covering results also suggest orbit-sharing constructions, but the sparse non-crossing graph is the more direct and testable ML contribution.

Ideas from this paper

Unverified 2026

Non-crossing d-arc sparse attention

Represent sequence positions as vertices on a circle and use a maximal family of mutually non-crossing d-arcs to define the allowed attention interactions. Rotate the family by d positions, or use several phase-shifted families across successive layers, so each layer has only O(N) edges but repeated layers propagate information over long distances. The geometric compatibility rule replaces arbitrary local-window or hand-designed sparse masks with a structured family whose maximality gives broad…

Useful5/10
Difficulty4/10
Novelty7/10
Paper: Higher cluster tilting objects in locally finite triangulated categories arXiv:2607.19916