JD-BP: A Joint-Decision Generative Framework for Auto-Bidding and Pricing

2026-04-07

AI Summary

This work addresses the suboptimal bidding strategies and inefficient ad allocation in automated bidding caused by prediction errors and feedback delays. To this end, we propose JD-BP, a joint decision-making framework that unifies bidding and pricing within a single model for the first time. Our approach introduces an additive pricing correction term and a memoryless Return-to-Go mechanism, further enhanced by trajectory augmentation and energy-based Direct Preference Optimization (EB-DPO) to improve collaborative learning. The method supports plug-and-play deployment and achieves state-of-the-art performance on AuctionNet in offline evaluations. Online A/B tests on JD.com demonstrate a 4.70% increase in ad revenue and a 6.48% reduction in target cost.

Abstract

Auto-bidding services optimize real-time bidding strategies for advertisers under key performance indicator (KPI) constraints such as target return on investment and budget. However, uncertainties such as model prediction errors and feedback latency can cause bidding strategies to deviate from ex-post optimality, leading to inefficient allocation. To address this issue, we propose JD-BP, a Joint generative Decision framework for Bidding and Pricing. Unlike prior methods, JD-BP jointly outputs a bid value and a pricing correction term that acts additively with the payment rule such as GSP. To mitigate adverse effects of historical constraint violations, we design a memory-less Return-to-Go that encourages future value maximizing of bidding actions while the cumulated bias is handled by the pricing correction. Moreover, a trajectory augmentation algorithm is proposed to generate joint bidding-pricing trajectories from a (possibly arbitrary) base bidding policy, enabling efficient plug-and-play deployment of our algorithm from existing RL/generative bidding models. Finally, we employ an Energy-Based Direct Preference Optimization method in conjunction with a cross-attention module to enhance the joint learning performance of bidding and pricing correction. Offline experiments on the AuctionNet dataset demonstrate that JD-BP achieves state-of-the-art performance. Online A/B tests at JD.com confirm its practical effectiveness, showing a 4.70% increase in ad revenue and a 6.48% improvement in target cost.

Problem

• auto-bidding
• pricing
• uncertainty
• KPI constraints
• inefficient allocation

Innovation

• Joint Decision Framework
• Pricing Correction
• Return-to-Go
• Trajectory Augmentation
• Energy-Based Direct Preference Optimization