Reasoning Primitives in Hybrid and Non-Hybrid LLMs

ArXi:2604.21454v1 Announce Type: new Reasoning in large language models is often treated as a monolithic capability, but its observed gains may arise from basic operations. We study reasoning through two such primitives, recall and state-tracking, and ask whether hybrid architectures that combine attention-based retrieval with recurrent state updates are better suited than attention-only models for tasks that jointly require both.