Taipan: Efficient and Expressive State Space Language Models with Selective Attention
This addresses the problem of long-context modeling for NLP applications, offering a promising solution that balances efficiency and performance, though it is incremental as it builds on existing SSMs and attention mechanisms.
The paper tackles the challenge of efficient long-context language modeling by introducing Taipan, a hybrid architecture combining Mamba-2 with Selective Attention Layers, which extends accurate predictions to context lengths of up to 1 million tokens while preserving computational efficiency.
Efficient long-context language modeling remains a significant challenge in Natural Language Processing (NLP). While Transformers dominate language tasks, they struggle with long sequences due to quadratic computational complexity in training and linearly scaling memory costs during inference. Recent State Space Models (SSMs) such as Mamba offer alternatives with constant memory usage, but they underperform in tasks requiring extensive in-context retrieval. We introduce Taipan, a novel hybrid architecture that combines Mamba-2 with Selective Attention Layers (SALs). These SALs identify tokens requiring long-range interactions, remove less important features, and then augment their representations using the attention module. This approach balances Mamba's efficiency with Transformer-like performance in memory-intensive tasks. By constraining the attention budget, Taipan extends accurate predictions to context lengths of up to 1 million tokens while preserving computational efficiency. Our experiments demonstrate Taipan's superior performance across various scales and tasks, offering a promising solution for efficient long-context language modeling.