xLSTM-Mixer integrates temporal, joint, and multi-view mixing with xLSTM blocks, achieving consistent SOTA performance in a wide range of long-term forecasting tasks.

GC-xLSTM jointly optimizes a sparse feature selector and predictive xLSTM to uncover a graph of Granger Causal dependencies

We introduce TiledFlashLinearAttention a faster kernel algorithm for Linear RNNs and mLSTMs by improved Sequence Parallelism.