Author

Bo Liu

h-index 9 View on Semantic Scholar

We have 2 of 22 papers

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.

Preprint Aug 2026

Maglev: Sliding Recurrent Memory

A recurrent Transformer architecture with fixed-size memory that generalizes sliding-window attention while remaining parallelizable during training and improves validation loss and downstream pretraining benchmarks over sliding-window and latent recurrent transformer baselines.

Bo Liu, Qiang Liu · 2 citations

#artificial intelligence Preprint Sep 2026

GradLev: Token-Parallel Test-Time Training Via Costate Prediction

It is observed that, given layer inputs and activation gradients (costates), online gradient descent admits exact parallel scans for both forward evaluation and reverse backpropagation.

Bo Liu, Qiang Liu · 0 citations