Sign inGet started
Paper · arXiv

On attention heads and bilinear forms

We study the symmetric and antisymmetric parts of bilinear forms in the attention heads of trained large language models. We introduce an orthogonally invariant profile map from real bilinear forms to a three-dimensional simplex and observe that profiles of trained bilinear forms accumulate near profiles of rank-one bilinear forms. We prove that the symmetric part of a bilinear form in an attention head is the sum of a hyperbolic form and a zero form for a Zariski-dense subset of query-key matrices.

AuthorsAndrew O'Desky
Primary categorycs.LG
Published19 September 2026
Last updated at source19 September 2026
cs.LGmath.SPstat.ML
Embed this entry
<iframe src="https://silytics.si/embed/item/ef4b2e1b-80b0-42f8-a861-b97ad57ed1cc" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Free, no tracking. More about feeds and embeds.

Silytics shows public metadata from arXiv and links to the original. Information may be incomplete or out of date. Last refreshed 17 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.