Huan Ren

LaRec: Unleashing LLM-based Latent Reasoning for Generative Recommendation

This work proposes LaRec, an efficient generative recommendation framework designed to unleash the potential of latent reasoning in LLMs by designing Latent Pre-training that empowers LLMs with latent reasoning capabilities by providing rich supervisory signals to the latent space reasoning via step-level alignment and process direction alignment.