Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2510.04800 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866911609688948736 |
|---|---|
| author | Bae, Sangmin Acun, Bilge Lin, Chien-Yu Habeeb, Haroun Kim, Seungyeon Luo, Liang Wang, Junjie Wu, Carole-Jean |
| author_facet | Bae, Sangmin Acun, Bilge Lin, Chien-Yu Habeeb, Haroun Kim, Seungyeon Luo, Liang Wang, Junjie Wu, Carole-Jean |
| contents | Recent progress in large language models demonstrates that hybrid architectures--combining self-attention mechanisms with structured state space models like Mamba--can achieve a compelling balance between modeling quality and computational efficiency, particularly for long-context tasks. While these hybrid models show promising performance, systematic comparisons of hybridization strategies and analyses on the key factors behind their effectiveness have not been clearly shared to the community. In this work, we present a holistic evaluation of hybrid architectures based on inter-layer (sequential) or intra-layer (parallel) fusion. We comprehensively evaluate these designs across multiple dimensions: language modeling and downstream task performance, long-context capabilities, scaling analysis, and training and inference efficiency. By investigating the core characteristics of their computational primitive, we identify the most critical elements for each hybridization strategy and further propose optimal design recipes for hybrid models. Our comprehensive analysis provides practical guidance and valuable insights for developing hybrid language models, facilitating the optimization of architectural configurations. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_04800 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Hybrid Architectures for Language Models: Systematic Analysis and Design Insights Bae, Sangmin Acun, Bilge Lin, Chien-Yu Habeeb, Haroun Kim, Seungyeon Luo, Liang Wang, Junjie Wu, Carole-Jean Computation and Language Recent progress in large language models demonstrates that hybrid architectures--combining self-attention mechanisms with structured state space models like Mamba--can achieve a compelling balance between modeling quality and computational efficiency, particularly for long-context tasks. While these hybrid models show promising performance, systematic comparisons of hybridization strategies and analyses on the key factors behind their effectiveness have not been clearly shared to the community. In this work, we present a holistic evaluation of hybrid architectures based on inter-layer (sequential) or intra-layer (parallel) fusion. We comprehensively evaluate these designs across multiple dimensions: language modeling and downstream task performance, long-context capabilities, scaling analysis, and training and inference efficiency. By investigating the core characteristics of their computational primitive, we identify the most critical elements for each hybridization strategy and further propose optimal design recipes for hybrid models. Our comprehensive analysis provides practical guidance and valuable insights for developing hybrid language models, facilitating the optimization of architectural configurations. |
| title | Hybrid Architectures for Language Models: Systematic Analysis and Design Insights |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2510.04800 |