发表机构
Institute for Intelligent Networked Systems (INSI); Northeastern University London（智能网络系统研究所; 伦敦东北大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
该研究提出TSDS框架，结合收敛探测器与困惑度延迟规则，经LTT流程校准后，可在边缘LLM智能体上减少43%-73%的思考计算量，同时保证奖励与云端调用率，在多类基准任务上表现优异。
AI 中文摘要
遵循ReAct范式的大语言模型智能体有望成为复杂多步任务的有力支撑，包括多跳问答、代码生成以及物理AI系统的控制。然而，当部署在边缘设备时，它们必须严格管理推理预算，同时保持可靠性，仅在本地不确定性过高无法安全行动时才延迟至云端模型。我们提出Think Short, Defer Smart（TSDS）框架，该框架协同集成了轻量收敛探测器（一旦预期行动稳定即终止设备端推理）与基于困惑度的延迟规则（将不确定行动升级至云端模型）。两种机制均通过多目标的Learn-Then-Test（LTT）流程在端到端回合轨迹上联合校准，为预期回合奖励与云端调用率提供有限样本保证。我们在四个ReAct基准上评估TSDS，涵盖算术推理（GSM8K）、多跳问答（HotpotQA）、代码生成（MBPP）及多步具身规划（家用机器人），并与仅思维校准、仅校准延迟的独立基线对比。TSDS在HotpotQA、MBPP及家用机器人任务上，相比仅延迟基线将每回合思考计算量降低43%-73%，同时保持经认证的奖励与云端调用率保证。
英文摘要
LLM agents following the ReAct paradigm are promising enablers of complex multi-step tasks, including multi-hop question answering, code generation, and control of physical AI systems. Yet, when deployed at the edge, they must tightly manage their reasoning budget while remaining reliable and deferring to a cloud-side model only when local uncertainty is too high to act safely. We propose Think Short, Defer Smart (TSDS), a framework that synergistically integrates a lightweight convergence probe, which halts on-device reasoning once the intended action has stabilized, with a perplexity-based deferral rule that escalates uncertain actions to a cloud-side model. Both mechanisms are jointly calibrated on end-to-end episode trajectories via a multi-objective Learn-Then-Test (LTT) procedure, providing simultaneous finite-sample guarantees on expected episode reward and cloud-call rate. We evaluate TSDS on four ReAct benchmarks spanning arithmetic reasoning (GSM8K), multi-hop question answering (HotpotQA), code generation (MBPP), and multi-step embodied planning (household robot), and compare against thought-calibration-only and calibrated-deferral-only standalone baselines. TSDS reduces per-episode thinking compute by 43%-65% over deferral-only baselines across HotpotQA, MBPP, and the household robot task, while maintaining certified reward and cloud-call rate guarantees.
