短く考え、賢く委譲し、行動を繰り返す：エッジLLMエージェントのための校正済み推論と不確実性認識型委譲
Think Short, Defer Smart, Act, and Repeat: Calibrated Reasoning and Uncertainty-Aware Deferral for Edge LLM Agents
エッジで動作するLLMエージェントの推論コストを削減しつつ信頼性を保つため、行動が安定したら推論を止めるプローブと、不確実な行動をクラウドに委譲するルールを組み合わせたフレームワークTSDSを提案し、4つのベンチマークで効果を検証した。
著者: Amirmohammad Farzaneh, Osvaldo Simeone
分類: stat.ML, cs.AI, cs.IT, cs.LG
原文アブストラクト
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%-73% over deferral-only baselines across HotpotQA, MBPP, and the household robot task, while maintaining certified reward and cloud-call rate guarantees.