Reproduce “AdKnob: Ad Intensity Control and Labeling for LLM-Native Advertising” 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
Woo Jae Kim, Seongho Keum, Joonsung Jeon, Suhyeon Ha, Sooel Son, Sung-eui Yoon
AdKnob lets providers control ad intensity in assistant answers and label promotional sentences with a fine-tuned model. It adds an intensity token and uses attention rollout to trace sentences to advertiser inputs. In Qwen3-4B, labeling reaches F1 0.76 versus 0.68 for BERTScore, with labeling time of 3.3 ms versus 1,785 ms. Results assume providers fine-tune models and access attention weights.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.