Memory improves proactive assistant timing and content decisions
ProPerAgent: Toward Memory-Driven Proactive Assistance
Computation and Language
Summary
Proactive assistants try to help you before you ask for something, but figuring out when and what to say is hard. The authors created a big test called ProPerMem that has long personal memories and asks AI to decide if and when to help. Many existing AI models struggle to correctly decide both the right information and the right timing. They designed ProPerAgent, which uses a smart memory system and a small learning step to improve both content accuracy and timing of the assistant’s interventions.
What this means in practice
- •For personal assistant developers: Create AI that better decides when and what proactive help to offer using extensive personal memory data.
- •For smart home device engineers: Design smart home assistants that reduce unnecessary interruptions by gating memory-based triggers with reinforcement learning.
Authors
Shenglai Zeng, Xiao Yang, Tiffany Ma, Xinyuan Zhang, Nanshu Wang, Deren Lei, Kai Sun, Xilun Chen, Tammy Stark, Xianhui Zhu, Yuetai Li, Rulin Shao, Tianxin Wei, Rakesh Wanga, Anuj Kumar, Jiliang Tang, Wen-tau Yih, Xin Luna Dong
Abstract
As assistants move from reactive answering toward proactive intervention, memory changes role: it no longer only supplies the answer to a query, it determines whether and when to speak at all. Prior work in this setting hands the system the relevant fact in context; we study the regime that makes the problem hard, where the evidence justifying an intervention must be mined from a long-horizon personal memory. We introduce ProPerMem, a benchmark for memory-driven proactive assistance (MPA), pairing egocentric video with year-long, distractor-rich user memories (~618K tokens on average), human-verified trigger windows, and a protocol that scores content and timing jointly. Every baseline we evaluate performs poorly (at most 15%), including frontier models given summarized or retrieval-augmented memory. A controlled study explains why: with oracle evidence delivered inside the trigger window GPT-5.3-Chat reaches 96%; letting it choose the timing drops it to 38%, and replacing oracle evidence with retrieval drops it to 15%. Content and timing are separately necessary, and missing either is catastrophic. Guided by this, we build ProPerAgent (Proactive Personalized Agent) with a plug-and-play harness that surfaces memory from a graph-organized store and gates it down to what is worth interrupting for, plus a lightweight RL stage that curbs early and repeated firing. Without any training, ProPerAgent lifts strict accuracy from below 10% to 37.5% on Qwen3.8-27B and 38% on GPT-5.3-Chat; RL post-training reaches 42.4%.