arXiv cs.CL· 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·· 12 小时前AI 评分42
ProPerAgent：面向记忆驱动主动式辅助的智能体
ProPerAgent: Toward Memory-Driven Proactive Assistance
AI 导读
研究者提出 ProPerMem 基准与 ProPerAgent，用于从长时个人记忆中挖掘干预依据并判断是否、何时主动介入。ProPerMem 平均约 618K tokens，现有基线最高仅 15%，GPT-5.3-Chat 在给定 oracle 证据时达 96%，自行决定时机降至 38%。
正文 · 原文
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%.
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.22671 [cs.CL] |
| (or arXiv:2610.22671v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.22671 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Shenglai Zeng [view email]
[v1]
Thu, 15 Oct 2026 21:40:17 UTC (12,296 KB)
来源：arXiv cs.CL · arxiv.org