AdsWorldEngine: AI Framework for Conversational Advertising
A recent paper published on arXiv (ID: 2608.13833) presents AdsWorldEngine, a framework aimed at enhancing conversational advertising. This innovative system tackles the challenge of deducing hidden commercial intent from user interactions, including queries, assistant replies, and dialogue history, by delivering relevant advertisements during multi-turn conversations. AdsWorldEngine features three main elements: an Opportunity Gate that decides when to display ads, an Orchestrator that formulates commercial intents, activates advertising tools, and creates a top-3 ad list, and an Evaluator that assesses ads for offline optimization. The key advancement lies in an iterative training method for the Orchestrator, which begins with supervised fine-tuning followed by agentic reinforcement learning, refining the process through high- and low-reward rollouts. This paper falls under the category of cross-type announcements and can be accessed via the provided URL.
Key facts
- AdsWorldEngine is an agentic framework for conversational advertising.
- It uses an Opportunity Gate to decide whether to show ads.
- The Orchestrator generates commercial intents and calls advertising tools.
- The Evaluator scores ads for offline optimization.
- Training involves supervised fine-tuning and agentic reinforcement learning.
- The framework constructs a top-3 ad slate.
- The paper is on arXiv with ID 2608.13833.
- The announcement type is cross.
Entities
Institutions
- arXiv