Expert-Guided Diffusion Planner for Auto-bidding

2025-08-12

AI Summary

To address insufficient structural modeling and low temporal efficiency in generative automated bidding, this paper proposes an expert-trajectory-guided diffusion planning method. The approach builds upon a conditional diffusion model framework, integrating expert demonstration trajectories as conditioning signals to enhance personalized structural modeling of optimal bidding sequences. Its key innovations are: (1) leveraging expert trajectories to explicitly guide the generation of structurally coherent, long-horizon bidding policies; and (2) introducing a non-Markovian modeling paradigm coupled with a leapfrog denoising sampling scheme, thereby circumventing the t-step autoregressive latency bottleneck inherent in conventional sequential models. Experimental results demonstrate that the method achieves high-quality, low-latency generation of extended bidding sequences. Offline evaluations confirm its effectiveness, while online A/B tests show statistically significant improvements—+11.29% in conversion rate and +12.35% in advertiser revenue.

Technology Category

• Search and Optimization: Learning to Search
• Planning, Routing, and Scheduling: Planning with Language Models
• Machine Learning: Online Learning & Bandits

Application Category

• Economics, Online Markets and Human Computation: Uses of LLMs and GenAI for marketplace design, bidding, and strategic interactions
• User Modeling, Personalization and Recommendation: User modeling for targeted and personalized online advertising
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Auto-bidding is extensively applied in advertising systems, serving a multitude of advertisers. Generative bidding is gradually gaining traction due to its robust planning capabilities and generalizability. In contrast to traditional reinforcement learning-based bidding, generative bidding does not rely on the Markov Decision Process (MDP) exhibiting superior planning capabilities in long-horizon scenarios. Conditional diffusion modeling approaches have demonstrated significant potential in the realm of auto-bidding. However, relying solely on return as the optimality condition is weak to guarantee the generation of genuinely optimal decision sequences, lacking personalized structural information. Moreover, diffusion models' t-step autoregressive generation mechanism inherently carries timeliness risks. To address these issues, we propose a novel conditional diffusion modeling method based on expert trajectory guidance combined with a skip-step sampling strategy to enhance generation efficiency. We have validated the effectiveness of this approach through extensive offline experiments and achieved statistically significant results in online A/B testing, achieving an increase of 11.29% in conversion and a 12.35% in revenue compared with the baseline.

Problem

• Enhancing auto-bidding decision sequences with expert-guided diffusion
• Addressing timeliness risks in diffusion-based generative bidding
• Improving auto-bidding performance via skip-step sampling strategy

Innovation

• Expert-guided diffusion modeling for auto-bidding
• Skip-step sampling to boost efficiency
• Combines expert trajectories with diffusion planning