Newest papers

Tandem: Serving Hybrid Linear-Attention GUI Agents on Mobile NPUs

Zhuomin Chen, Jingchao Ni, Xu Zheng, Yong Chen, Mo Sha, Wei Cheng, Dongsheng Luo · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.26660 · cited by 0 works (OpenAlex) · open-access PDF

Portuguese

As unidades de processamento neural móvel (NPUs) permitem a execução local, mantendo o conteúdo da tela no telefone, mas a latência do modelo se acumula com cada ação. Os modelos de atenção linear híbrida reduzem os requisitos de memória, mas a reutilização do seu estado de matriz requer pontos de verificação nas posições de entrada apropriadas. As solicitações do agente retêm as instruções, mas reescrevem o histórico recente, de modo que as políticas do ponto de verificação projetadas para a continuação da conversação podem perder prefixos reutilizáveis. A retenção de estados de rollback de texto para a decodificação especulativa também pode mover o processamento prompt da unidade de matriz da NPU, acrescentando o overhead que supera as economias de decodificação nas saídas de ação curtas

Machine translation (Meta M2M100); the English original follows.

Abstract

Smartphone GUI agents complete tasks in mobile apps by repeatedly reading the screen and asking a model for the next action. Mobile neural processing units (NPUs) enable local execution, keeping screen contents on the phone, but model latency accumulates with every action. Hybrid linear attention models reduce memory requirements, yet reusing their recurrent state requires checkpoints at the appropriate input positions. Agent requests retain instructions but rewrite recent history, so checkpoint policies designed for chat continuation can miss reusable prefixes. Retaining rollback states for speculative decoding can also move prompt processing off the NPU's matrix unit, adding overhead that outweighs decoding savings on short action outputs. A loaded model can further hold the NPU awake between tasks. We present Tandem, a runtime for hybrid GUI agents on mobile NPUs. Tandem saves recurrent state where consecutive prompts diverge, allowing later requests to reuse previously processed context. It keeps most prefill on the matrix unit and produces rollback states through a short per-token tail. Its verification kernel reads each weight block once for all proposed tokens. Between request bursts, Tandem releases NPU power votes without unloading the model. We implement Tandem in llama.cpp and evaluate it on recorded agent interactions across two Snapdragon generations. On a Snapdragon 8 Elite, Tandem reduces per-step latency by 30% for text-based screen descriptions and 16% for screenshots relative to llama.cpp's fastest NPU configuration, reduces energy per text step by 33%, and lowers standby power with the model loaded from 0.76 to 0.22 W.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Latency (audio), Operating system, Decoding methods, Reuse, Rollback

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange