Fetching from the wire…
01
02
03
04
05
06
07
08
Research2026-06-22 · source-backed
Przemyslaw Musialski reframes a token as an element of a matrix Lie group instead of a feature vector, scoring attention via a parameter-free negative squared algebra norm of the relative pose. On sequence-completion over SE(2), SO(3), and Aff(2) it matches learned MLP kernels with 50–80x fewer score parameters, while vector-token baselines violate invariance by 5–12 orders of magnitude. It extends equivariant attention to non-compact, non-abelian affine groups earlier methods couldn't handle. Niche, but directly relevant to robotics, pose, and geometric ML.
Each link below shares sources, entities, or timing with this story.
Shared entity: MLP / Same source domain / What happened next / Tension
Both cover MLP; reported by the same outlet (arxiv.org); picks up the MLP thread on 2026-07-30.
Shared entity: Niche / Same source domain / What happened next / Downstream implication
Both cover Niche; reported by the same outlet (arxiv.org); picks up the Niche thread on 2026-06-23.
Shared entity: MLP / Same source domain / Earlier coverage / Downstream implication
Both cover MLP; reported by the same outlet (arxiv.org); earlier MLP coverage from 2026-03-12.
Same source domain / Shared topic / Tension
Reported by the same outlet (arxiv.org); overlapping topics (attention, extend); pushes against this story (but).
Shared entity: Niche / Earlier coverage / Tension
Both cover Niche; earlier Niche coverage from 2026-06-19; pushes against this story (but).
Shared entity: MLP / What happened next
Both cover MLP; picks up the MLP thread on 2026-08-11.
Both cover MLP; picks up the MLP thread on 2026-07-25.
Same source domain / Shared topic
Reported by the same outlet (arxiv.org); overlapping topics (baselin, directly).