AI RESEARCH

CoMem: Context Management with A Decoupled Long-Context Model

arXiv CS.LG

ArXi:2605.30842v1 Announce Type: new Context management enables agentic models to solve long-horizon tasks through iterative summarization of previous interaction histories. However, this process typically incurs substantial decoding overhead for the extra summarization tokens, which significantly affect the end-to-end response latency at deployment. In this paper, we