most citedMemory-Augmented Transformers: A Systematic Review from Neuroscience Principles to Enhanced Model Architectures

1 citations · 1 across the 4 of their papers we have counts for

collaborators
Showing cs.AIShow all

1 paper · 1 filter

cs.AI2026

AllMem: A Memory-centric Recipe for Efficient Long-context Modeling

Ziming Wang, Xiang Wang, Kailong Peng +5

Large Language Models (LLMs) encounter significant performance bottlenecks in long-sequence tasks due to the computational complexity and memory overhead inherent in the self-atten…