Reproduce “Lightweight Auto-bidding based on Traffic Prediction in Live 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
Bo Yang, Ruixuan Luo, Junqi Jin, Han Zhu
The BiCB algorithm offers a lightweight, real-time auto-bidding solution for live advertising by integrating Lagrangian dual analysis with statistical aggregate traffic prediction. Deployed at Alibaba, it increased Gross Merchandise Volume by 4.28% and reduced CPC constraint violations by 6.25% compared to existing baselines.
Key insights
- Optimal auto-bidding in live advertising can be efficiently achieved by combining classical Lagrangian duality with lightweight aggregate future traffic prediction, bypassing complex sequence modeling.
- The core auto-bidding challenge, when viewed as a knapsack-like problem, benefits more from understanding overall traffic distribution than precise per-impression sequencing.
- Expanding the problem formulation to include both upper and lower bound constraints for KPIs offers advertisers more granular control over traffic quality and budget adherence.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.