Reproduce “MemoProAgent: Toward Memory-Driven Proactive Assistance” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Shenglai Zeng, Xiao Yang, Tiffany Ma, Xinyuan Zhang, Nanshu Wang, Deren Lei, Kai Sun, Xilun Chen and 10 more
MemoProBench tests proactive assistance using egocentric video and long synthetic histories. In a 150-clip study, GPT-5.3-Chat reached 96% Perfect with oracle evidence and timing, but 38% when choosing timing and 15.3% with retrieved evidence. MemoProAgent reached 37.5% on 800 clips, rising to 42.4% with reinforcement learning. Histories are synthetic; each clip has one trigger window.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.