Reproduce “The Web Doesn't Sit Still: Adversarial Self-Evolving Attacks on Search 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
Geyuan Wang, Shunyu Liu, Siyuan Liang, Chenyang Lyu, Dacheng Tao
The authors introduce CREO, a framework that evolves search-agent poisoning strategies from successes and failures while updating defensive prompts with known attack patterns. In controlled multi-hop question-answering tests, CREO achieved lower strict accuracy than static baselines in most model–dataset combinations. For GLM-5.1 on MuSiQue, accuracy was 16% versus 41% for AuthChain in Evo; live search transfer remains untested.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.