Reproduce “LLMAdBench: A Human Preference Benchmark for Advertising in LLM Responses” 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
Rui Ai, Yuqing Liu, Sitao Qiu, Yun Qiao, Yuhan Wang, Jessica Xiwen Wang, Yiqi Yang, Lihong Huang and 9 more
LLMAdBench compares human preferences for four positions of matched ads in otherwise fixed assistant answers, with and without sponsorship labels. Its renderer can locally reword ads while preserving their entity and core message. Late placement leads overall when unlabeled (0.352), while End leads when labeled (0.364); these model-based scores reflect benchmark judgments, not real-world clicks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.