Below is a structured breakdown of the core paper and the relevant "fix" or evolution regarding Auto-Regressive capabilities. The foundational paper that established this architecture is: Hdking Guru Site
While there is no single widely cited paper titled "justARKAN" yet (as this is an emerging term in the open-source community), the architecture refers to replacing the MLP layers inside standard Auto-Regressive models (like LSTMs, GRUs, or Transformer feed-forward layers) with KAN layers. Regjistri I Gjendjes Civile: Nentor 2008 Ver 14 Best
Since this is a very recent development in the field of Deep Learning (building upon the introduction of Kolmogorov-Arnold Networks in May 2024), "justARKAN fix" usually refers to specific modifications or the original implementation that "fixes" the complexities of Multi-Layer Perceptrons (MLPs) by replacing them with spline-based KAN layers.