Reproduce “VirusCascade: Hijacking Collaborative Reflection in LLM-Powered Recommender 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

Yurong Hao, Wen Zhou, Guowei Guan, Tiantong Wu, Fuyao Zhang, Wei Yang Bryan Lim

The authors analyze reflection-based recommender agents and introduce VirusCascade, an attack combining rewritten item profiles with interaction paths through popular items to encourage injected preferences to persist and spread. Across six simulated configurations, the target entered the top 20 for 0.384 of users, versus 0.199 for strongest baselines; results use sampled users and items, not real platforms.

Summary by alphaXiv.

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