4 papers · 1 filter
Complex KDA: Understanding and Enhancing the Expressivity of Kimi Delta Attention
Julien Siems, Riccardo Grazzi, Korbinian Pöppel +8
Linear RNNs based on the delta-rule enable efficient sequence modeling, but their linear updates with a low-rank correction constrain their expressivity. Prior work has shown that…
xLSTM 7B: A Recurrent LLM for Fast and Efficient Inference
Maximilian Beck, Korbinian Pöppel, Phillip Lippe +5
Recent breakthroughs in solving reasoning, math and coding problems with Large Language Models (LLMs) have been enabled by investing substantial computation budgets at inference ti…
Tiled Flash Linear Attention: More Efficient Linear RNN and xLSTM Kernels
Maximilian Beck, Korbinian Pöppel, Phillip Lippe +1
Linear RNNs with gating recently demonstrated competitive performance compared to Transformers in language modeling. Although their linear compute scaling in sequence length offers…
FlashRNN: I/O-Aware Optimization of Traditional RNNs on modern hardware
Korbinian Pöppel, Maximilian Beck, Sepp Hochreiter
While Transformers and other sequence-parallelizable neural network architectures seem like the current state of the art in sequence modeling, they specifically lack state-tracking…