Search papers, labs, and topics across Lattice.
The paper introduces ProAct, a proactive agent architecture that utilizes idle-time compute to anticipate user needs by analyzing dialogue history and persistent memory. This allows the agent to pre-emptively acquire information and prepare evidence, addressing knowledge gaps before user queries. Evaluated on the new ProActEval benchmark, ProAct demonstrates significant improvements over reactive agents, including reduced turns, user effort, and hallucination rates, along with state-of-the-art reflective accuracy on MemBench.
Stop wasting idle compute: ProAct agents anticipate user needs and proactively gather information, slashing task completion time and hallucinations.
While AI agents demonstrate remarkable capabilities in reasoning and tool use, they remain fundamentally reactive: they compute responses only after explicit user prompts. This paradigm ignores a critical opportunity: the idle time between interactions is largely wasted, leaving agents unable to prepare for future user needs. To bridge this gap, we introduce ProAct, a proactive agent architecture that leverages idle-time compute to anticipate and fulfill likely upcoming user needs. By analyzing evolving dialogue history together with persistent memory, ProAct predicts upcoming needs and iteratively acquires information, allowing the agent to resolve knowledge gaps and prepare evidence before the user initiates a query. To rigorously evaluate proactive capabilities, we also introduce ProActEval, a comprehensive benchmark comprising 200 scenarios across 40 domains, featuring predictable need chains and diverse user cognitive profiles. Empirical results demonstrate significant advantages over reactive baselines. ProAct accelerates task completion by reducing required turns by 14.8%, decreases user effort by 11.7%, and cuts hallucination rates by 28.1% on ProActEval. Furthermore, MemBench evaluations confirm that ProAct achieves state-of-the-art reflective accuracy, underscoring its sustained and robust performance.
