Free Lesson
Semantic IDs for Generative Recommender Systems
60 min
Dec 4, 2026 10:30 AM
By continuing, you agree to Maven's Terms and Privacy Policy.
What you'll learn
When generative recommenders make sense
Learn where foundation models can improve recommendations—and where traditional methods may still win.
How Semantic IDs represent items for LLMs
See how Semantic IDs turn catalog items into representations language models can generate and reason over.
How to train an LLM recommender
Walk through Semantic ID generation, fine-tuning, and optimization strategies for efficient training.
How to evaluate generative recommendations
Learn practical approaches for measuring the quality of LLM-based recommender systems.
Why this topic matters
Recsys is moving beyond separate retrieval and ranking models toward systems that can generate items directly. Semantic IDs make that possible by giving LLMs a vocabulary for products, content, and other catalog entities. Understanding this approach helps search and recommendation practitioners evaluate where generative models can outperform—or complement—traditional systems.
