Smarter reward signals improve product recommendation models

Original: Adaptive Loss Balancing for Noise-Robust GRPO in Generative Recommendation

Source: arxiv.org ↗