MAR: Efficient Large Language Models via Module-aware Architecture Refinement
This addresses energy efficiency issues in large language models, which is an incremental improvement for AI deployment in resource-constrained environments.
The paper tackles the high energy costs of large language models by proposing MAR, a framework that integrates state space models for linear-time sequence modeling and applies activation sparsification, reducing inference energy consumption while restoring performance compared to dense models.
Large Language Models (LLMs) excel across diverse domains but suffer from high energy costs due to quadratic attention and dense Feed-Forward Network (FFN) operations. To address these issues, we propose Module-aware Architecture Refinement (MAR), a two-stage framework that integrates State Space Models (SSMs) for linear-time sequence modeling and applies activation sparsification to reduce FFN costs. In addition, to mitigate low information density and temporal mismatch in integrating Spiking Neural Networks (SNNs) with SSMs, we design the Adaptive Ternary Multi-step Neuron (ATMN) and the Spike-aware Bidirectional Distillation Strategy (SBDS). Extensive experiments demonstrate that MAR effectively restores the performance of its dense counterpart under constrained resources while substantially reducing inference energy consumption. Furthermore, it outperforms efficient models of comparable or even larger scale, underscoring its potential for building efficient and practical LLMs.