Andrew Fairless, Ph.D.
blue-green banner
What I Read: LLM-RecSys Hybrid.
Posted on :: Tags: recommender, natural language processing, large language model, token, autoencoder, variational autoencoder, loss, embedding

https://eugeneyan.com//writing/semantic-ids/
Training an LLM-RecSys Hybrid for Steerable Recs with Semantic IDs
Eugene Yan
14 Sep 2025
"Instead of using random hash IDs for videos or songs or products, we can use semantically meaningful tokens that an LLM can natively understand.... I could simply chat with the model to steer its recommendations..."