{"page":"/blog/reinforcement-learning-bidding","modified":"2026-11-10","license":"https://creativecommons.org/licenses/by/4.0/","columns":["Organisation","What the public source describes","RL?","Source"],"rows":[["Alibaba","RL bidder deployed in e-commerce search auction; generative auto-bidder tested online","Yes","Zhao et al., 2018; Guo et al., 2024"],["DoorDash","Smart Fast Finish: adjusts start time and throttle rate from historical campaign data to stop overdelivery","No, tuned control","Garg et al., 9 Sep 2025"],["Google (paper authors)","Budget and return-on-spend pacing algorithms compared on data from a large advertising platform","No, dual-based pacing","Balseiro et al., 2023"],["Topsort","Throttling and discount pacing in its auction","Not stated","Topsort docs, accessed 9 Nov 2026"],["Yahoo (paper authors)","Smart Pacing learned from data in a real DSP","No, adaptive control","Xu et al., 2015"]],"sourceUrls":[["https://arxiv.org/abs/1803.00259","https://arxiv.org/abs/2405.16141"],["https://arxiv.org/abs/2509.07929"],["https://arxiv.org/abs/2302.08530"],["https://docs.topsort.com/knowledge-base/ad-server/auctions/pacing/"],["https://arxiv.org/abs/1506.05851"]]}