Reproduce “Beyond Sequences: Distilling Structured Decision Memory for LLM Recommendation” 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
Leikun Liang, Guoshuai Wang, Xingsheng He, Yushan Han, Yunyi Xuan, Xiaoxiao Xu, Lin Qu
MARI recommends items by retrieving compact, readable records of past purchase decisions and combining relevant records with recent behavior. Its offline process uses PHDD to create structured decision memories, which encode rationales and preference factors. On Difficult Choice Prediction, fine-tuned Qwen3-4B achieves 0.8868 Hit@1, versus 0.5158 for untuned DeepSeek-R1, under a small same-category candidate-set evaluation, not full-catalog recommendation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.