Researchers at the AI Center of Lomonosov Moscow State University have developed the RobustBid algorithm, which increases the stability of automated bidding systems in digital advertising. Such systems are used in advertising auctions, where algorithms themselves calculate how much to bid for an ad impression.
Autobidding relies on machine learning predictions: for example, the probability of a click on an ad and the probability of a target action after it. The problem is that such predictions can be wrong, and because of this, advertising campaigns lose effectiveness.
The RobustBid method takes into account the uncertainty of predictions when forming bids. This makes the system less sensitive to model errors and shows more stable results compared to basic automated bidding algorithms.
MSU notes that the development can be useful not only for advertising auctions but also for other tasks where decisions are made based on predictive models and it is important to take into account possible errors in estimates. The research results were presented at the international conference AAMAS 2026.
Read more on the topic:
- How Internet Algorithms Affect Children: Cyber Police Warns of the Danger of Recommendation Systems
- Deadlines Announced for Advertising on Telegram and YouTube
- Smartphone for watching ads started to be given out in Russia