Reproduce “MemTrial: Learning When to Trust Memory in LLM Portfolio Agents” 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
Guanghao Wu, Zhuo Cai, Shoujin Wang
MemTrial estimates a lesson’s contribution by comparing portfolio drafts with and without it on the same market date. It pools these noisy estimates across dates and trusts learned values only when they predict later contributions. In PlantedMem, it achieved 0.87% monthly utility, best among 15 methods; on real benchmarks, it stayed near equal weighting, without establishing that memory improves returns.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.