Author

Joe Logan

1 paper indexed here

We haven’t gathered this author’s papers yet. Follow them and we’ll fetch their work.

Not the right person? Other researchers publish under this name.

Jul 2026

Per-Token Fixed-Point Convergence in Depth-Recurrent Transformers

It is shown that reading it is directly readable and that reading it outperforms learning to predict it: a training-free rule that halts each token once its output stabilizes attains uniform depth-8 quality and matches uniform depth across the average-depth range.

Joe Logan · arXiv.org · 2 citations