2024/12/26 by Jingcheng Hu, Hu, Jingcheng, Houyi Li +14 · 1 voice · 2 citations
Computer Science · #Cognitive Science and Mapping #Computation and Language (cs.CL) #FOS: Computer and information sciences #Machine Learning (cs.LG) #cs.CL #cs.LG
paper · pdf · doi:10.48550/arxiv.2412.19255
openalex publication_date 2024/12/26 · arxiv published 2024/12/26 · arxiv updated 2025/01/14 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
We propose novel attention architectures, Multi-matrix Factorization Attention (MFA) and MFA-Key-Reuse (MFA-KR). Existing variants for standard Multi-Head Attention (MHA), including SOTA methods like MLA, fail to maintain as strong performance under stringent Key-Value cache (KV cache) constraints. MFA enhances model capacity by efficiently scaling up both the number and dimension of attention heads through low-rank matrix factorization in the Query-Key (QK) circuit. Extending MFA, MFA-KR further reduces memory requirements by repurposing the key cache as value through value projection re-parameterization. MFA's design enables strong model capacity when working under tight KV cache budget, while MFA-KR is suitable for even harsher KV cache limits with minor performance trade-off. Notably, in our extensive and large-scale experiments, the proposed architecture outperforms MLA and performs comparably to MHA, while reducing KV cache usage by up to 56% and 93.7%, respectively.