الملخص
KV-cache quantization and linear attention are two representative approaches to tackling the storage and computational costs of Transformers. KV-cache quantization compresses individual KV entries into discrete codes but retains all entries, whereas linear attention recurrently aggregates multiple historical KV contributions into a fixed-size continuous state but can introduce interference. This contrast raises the question of whether per-KV compression and multi-KV aggregation can be bridged within a single mechanism for efficient attention. We identify RAM-Net as such a bridge through soft assignments over a discrete address space. These assignments determine recurrent updates to the continuous slot state associated with each address. Under a restricted RAM-Net construction, we prove that soft address assignments extend hard quantized matching to a separable read-write overlap that locally approximates full-attention similarity and supports recurrent aggregation. These connections further enable Transformer-to-RAM-Net weight migration through a new path based on a soft-quantized intermediate construction. Across nine pretrained Transformer models from 0.3B to 7B parameters, RAM-Net recovers an average of 87.1% of the teachers' accuracy gains over random guessing across six commonsense and knowledge tasks using only a 500M-token budget per model.
الكلمات المفتاحية
الموضوع
بيانات النشر
- المجلة
- غير متاح
- وصول مفتوح
- وصول مفتوح أخضر
اقتبس هذه المقالة
APA 7
Xiao, K., Dong, L., Li, H., & Xing, G. (2026). Bridging KV-Cache Quantization and Linear Attention: From Theory to Pretrained Weight Migration. https://omanscience.com/ar/articles/bridging-kv-cache-quantization-and-linear-attention-from-theory-to-pretrained-weight-migration
MLA 9
Xiao, Kaicheng, et al. "Bridging KV-Cache Quantization and Linear Attention: From Theory to Pretrained Weight Migration." https://omanscience.com/ar/articles/bridging-kv-cache-quantization-and-linear-attention-from-theory-to-pretrained-weight-migration.
شيكاغو (المؤلف–التاريخ)
Xiao, Kaicheng, Liran Dong, Haotian Li, and Guoliang Xing. 2026. "Bridging KV-Cache Quantization and Linear Attention: From Theory to Pretrained Weight Migration." https://omanscience.com/ar/articles/bridging-kv-cache-quantization-and-linear-attention-from-theory-to-pretrained-weight-migration.
هارفارد
Xiao, K., Dong, L., Li, H. and Xing, G. (2026) 'Bridging KV-Cache Quantization and Linear Attention: From Theory to Pretrained Weight Migration', Available at: https://omanscience.com/ar/articles/bridging-kv-cache-quantization-and-linear-attention-from-theory-to-pretrained-weight-migration.
فانكوفر
Xiao K, Dong L, Li H, Xing G. Bridging KV-Cache Quantization and Linear Attention: From Theory to Pretrained Weight Migration. https://omanscience.com/ar/articles/bridging-kv-cache-quantization-and-linear-attention-from-theory-to-pretrained-weight-migration
IEEE
K. Xiao, L. Dong, H. Li, and G. Xing, "Bridging KV-Cache Quantization and Linear Attention: From Theory to Pretrained Weight Migration," https://omanscience.com/ar/articles/bridging-kv-cache-quantization-and-linear-attention-from-theory-to-pretrained-weight-migration.