arXiv 2510.06828

Recurrence-Complete Frame-based Action Models

By Michael Keiblinger

Published 2025-10-08

Citation lineage

Review the prior work and downstream research connected to this paper.

In recent years, attention-like mechanisms have been used to great success in the space of large language models, unlocking scaling potential to a previously unthinkable extent. "Attention Is All You Need" famously claims RNN cells are not needed in conjunction with attention. We challenge this view. In this paper, we point to existing proofs that architectures with fully parallelizable forward or backward passes ca…

View the original paper on arXiv