paperGoogle Scholar·25 Oct 2026·X Zhang, L Wang, L Xiao
A stochastic-computing hardware accelerator for rotary position embedding in large language models
Stochastic-computing hardware accelerator for RoPE in LLMs.
Abstract
Integration, 2026 - Elsevier. Rotary position embedding (RoPE) encodes relative position in most large language models (LLMs), but in hardware it requires many trigonometric evaluations and vector …