Maximilian Beck

xLSTM Scaling Laws: Competitive Performance with Linear Time-Complexity

Short Window Attention Enables Long-Term Memorization

Tiled Flash Linear Attention: More Efficient Linear RNN and xLSTM Kernels

xLSTM 7B: A Recurrent LLM for Fast and Efficient Inference

PointMapPolicy: Structured Point Cloud Processing for Multi-Modal Imitation Learning

FlashRNN: I/O-Aware Optimization of Traditional RNNs on modern hardware

A Large Recurrent Action Model: xLSTM enables Fast Inference for Robotics Tasks

A Large Recurrent Action Model: xLSTM enables Fast Inference for Robotics Tasks

Vision-LSTM: xLSTM as Generic Vision Backbone