Reproduce “GuardAgent: A Memory-Augmented Agentic Framework for Adaptive Detection of Adversarial Promotions in LLM-Based 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

Shiyi Yang, Xinshu Li, Chen Wang, Xiwei Xu, Liming Zhu, Lina Yao

GuardAgent detects promotion attacks on item text in LLM-based recommenders by revising human-readable rules from labeled errors in batches and using retrieved examples. On Amazon Musical Instruments, it reports 0.92 F1 and 0.88 recall, versus 0.86 F1 and 0.79 recall for Retrieval-based ICL. It reports 100% detection of held-out DrunkAgent attacks with negligible false positives, but this covers constructed tests, not all future attacks.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.