HALO: Hindsight-Augmented Learning for Online Auto-Bidding

TL;DR

HALO is a novel online auto-bidding method that uses hindsight and B-spline representations to adapt efficiently across diverse budget and ROI constraints in real-time digital advertising auctions.

Contribution

HALO introduces a theoretically grounded hindsight mechanism and B-spline bid mapping to improve adaptability and generalization in multi-constraint auto-bidding environments.

Findings

Outperforms traditional methods in constraint handling.

Reduces constraint violations significantly.

Enhances GMV in industrial datasets.

Abstract

Digital advertising platforms operate millisecond-level auctions through Real-Time Bidding (RTB) systems, where advertisers compete for ad impressions through algorithmic bids. This dynamic mechanism enables precise audience targeting but introduces profound operational complexity due to advertiser heterogeneity: budgets and ROI targets span orders of magnitude across advertisers, from individual merchants to multinational brands. This diversity creates a demanding adaptation landscape for Multi-Constraint Bidding (MCB). Traditional auto-bidding solutions fail in this environment due to two critical flaws: 1) severe sample inefficiency, where failed explorations under specific constraints yield no transferable knowledge for new budget-ROI combinations, and 2) limited generalization under constraint shifts, as they ignore physical relationships between constraints and bidding coefficients. To address this, we propose HALO: Hindsight-Augmented Learning for Online Auto-Bidding. HALO introduces a theoretically grounded hindsight mechanism that repurposes all explorations into training data for arbitrary constraint configuration via trajectory reorientation. Further, it employs B-spline functional representation, enabling continuous, derivative-aware bid mapping across constraint spaces. HALO ensures robust adaptation even when budget/ROI requirements differ drastically from training scenarios. Industrial dataset evaluations demonstrate the superiority of HALO in handling multi-scale constraints, reducing constraint violations while improving GMV.