arXiv:2606.14192v1 Announce Type: new 
Abstract: Auto-bidding is a core component of real-time advertising systems, where decisions must optimize long-term performance under budget and cost constraints, while online exploration is prohibitively risky. Offline reinforcement learning and, more recently, Transformer-based sequence modeling have shown promise for learning bidding policies from logged data, but their unimodal and purely parametric formulations often collapse multiple effective bidding strategies into suboptimal averaged actions and perform unreliably under sparse or long-tail traffic. To mitigate these limitations, we propose DRIVE (Distributional and Retrieval-Augmented Bidding with Value Evaluation), a unified Transformer-based framework that decouples candidate action generation from decision making for offline auto-bidding. DRIVE combines distributional action modeling, retrieval-augmented candidate generation from high-quality historical decisions, and value-based evaluation to select the most promising bid at inference time. Extensive experiments on AuctionNet and additional offline reinforcement learning benchmarks demonstrate that DRIVE consistently improves bidding performance and generalizes well across multiple Transformer-based methods.

تم تقديم إطار عمل جديد يسمى DRIVE (المزايدة المعززة بالتوزيع وتقييم القيمة) لتحسين المزايدة التلقائية في أنظمة الإعلان في الوقت الحقيقي. يهدف هذا النموذج القائم على Transformer إلى تحسين الأداء على المدى الطويل تحت قيود الميزانية من خلال فصل توليد الإجراءات عن اتخاذ القرار، مما يعالج قيود استراتيجيات المزايدة التقليدية.

Se ha introducido un nuevo marco llamado DRIVE (Subasta Aumentada y Distribucional con Evaluación de Valor) para mejorar la auto-subasta en sistemas publicitarios en tiempo real. Este modelo basado en Transformer tiene como objetivo optimizar el rendimiento a largo plazo bajo restricciones presupuestarias al desacoplar la generación de acciones de la toma de decisiones, abordando así las limitaciones de las estrategias de subasta tradicionales.

Un nouveau cadre nommé DRIVE (Distributional and Retrieval-Augmented Bidding with Value Evaluation) a été introduit pour améliorer l'auto-enchère dans les systèmes publicitaires en temps réel. Ce modèle basé sur Transformer vise à optimiser la performance à long terme sous des contraintes budgétaires en dissociant la génération d'actions de la prise de décision, abordant ainsi les limitations des stratégies d'enchères traditionnelles.

A new framework named DRIVE (Distributional and Retrieval-Augmented Bidding with Value Evaluation) has been introduced to enhance auto-bidding in real-time advertising systems. This Transformer-based model aims to optimize long-term performance under budget constraints by decoupling action generation from decision-making, addressing the limitations of traditional bidding strategies.