Reproduce “Personal-Agent Mediated Recommendation with Cross-Platform User History” 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
Yu Xia, Jiangfan Zhang, Jun Xiao, Julian McAuley, Xiangjun Fan
PAMO trains a recommendation agent to mediate platform rankings using cross-platform history. It compares response rationales with full and masked histories to estimate dependence, then reallocates training advantage using recommendation outcomes and a platform-relative value floor. On MediateRec-OpenPlay, Qwen3-4B reaches HR@10 47.9% and NDCG@10 0.284 versus 47.4% and 0.267 for matched GRPO. Results cover candidate-conditioned ranking, not retrieval.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.