Block-recurrent Transformers target linear scaling
An article published on September 17, 2026 presents a recurrent, block-based Transformer architecture with complexity linear in sequence length.
Source: Block-Recurrent Transformers processam sequências com complexidade linear (arxiv.org). Text prepared with AI from this source.
What happened and what to do
Published on September 17, 2026, the article presents a Transformer layer applied recurrently across a sequence. Its recurrent cell processes blocks of tokens and combines self-attention with cross-attention; the article describes block recurrence as an alternative for long sequences with complexity linear in sequence length.
A company working with long documents or extended histories could evaluate this approach on its own data: compare quality, latency, and resource consumption with its current processing, and monitor results before deciding whether to adopt it. Implementation could include an evaluation pipeline and an API integrated with existing systems.
How the consultancy can help
Wendelmaques can diagnose the case’s requirements and exposure, define a technical evaluation with quality and cost criteria, and propose implementation and operation of a pipeline or API on the company’s infrastructure.
Next step
Send a short description of your case to receive a scoped proposal for diagnosis and implementation.
Let's assess your project
I reply with scope and a proposal.