C2:Cross learning module enhanced decision transformer with Constraint-aware loss for auto-bidding

2026-01-28

AI Summary

This work addresses the limitations of existing Decision Transformer models in automated bidding, which struggle to capture cross-sequence dependencies among states, actions, and returns and fail to effectively distinguish high-quality trajectories that satisfy budget and cost-per-acquisition (CPA) constraints. To overcome these challenges, the authors propose a Cross-Learning Block (CLB) to enhance multi-sequence cross-attention mechanisms and introduce a constraint-aware loss function (CL) to enable selective learning of high-value trajectories. Built upon the Decision Transformer architecture, offline experiments on the AuctionNet dataset demonstrate that the proposed method consistently outperforms current approaches across various budget settings, achieving performance gains of up to 3.2%. Ablation studies further confirm the synergistic effectiveness of the CLB and CL components.

Technology Category

• Constraint Satisfaction and Optimization: Constraint Learning and Acquisition
• Search and Optimization: Learning to Search
• Machine Learning: Online Learning & Bandits

Application Category

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

Abstract

Decision Transformer (DT) shows promise for generative auto-bidding by capturing temporal dependencies, but suffers from two critical limitations: insufficient cross-correlation modeling among state, action, and return-to-go (RTG) sequences, and indiscriminate learning of optimal/suboptimal behaviors. To address these, we propose C2, a novel framework enhancing DT with two core innovations: (1) a Cross Learning Block (CLB) via cross-attention to strengthen inter-sequence correlation modeling; (2) a Constraint-aware Loss (CL) incorporating budget and Cost-Per-Acquisition (CPA) constraints for selective learning of optimal trajectories. Extensive offline evaluations on the AuctionNet dataset demonstrate consistent performance gains (up to 3.2% over state-of-the-art method) across diverse budget settings; ablation studies verify the complementary synergy of CLB and CL, confirming C2's superiority in auto-bidding. The code for reproducing our results is available at: https://github.com/Dingjinren/C2.

Problem

• auto-bidding
• Decision Transformer
• cross-correlation modeling
• optimal behavior learning
• sequence modeling

Innovation

• Cross Learning Block
• Constraint-aware Loss
• Decision Transformer
• Auto-bidding
• Cross-attention