Work

Attention normalization and LayerNorm placement in graph Transformers: protocol and code

Year: 2026
Type: preprint
Abstract: Protocol (predictions P1-P7 with three amendments) and code of the study 'Attention normalization and LayerNorm placement limit which targets graph Transformers can extrapolate to larger graphs' (Yuqi Bai, Haoran Ma; manuscript under review). The protocol was written on 2026-09-30 in a local document and was not deposited with a third party at the time. This archive was created on 2026-11-13, after the study; it documents the content of the protocol, not when it was written.
Language: English
Open Access status: green
OpenAlex ID: W7225989946
Cites:
Cited by:
Related to: 10