MORL-A2C: Multi-Objective Reinforcement Learning Reranker for Optimizing Healthiness in Food Recommendation

arXiv·low signal

Recommendation systems that optimize purely for user preference can reinforce unhealthy dietary behavior; MORL-A2C is a multi-objective RL reranker that balances preference against nutritional health. It's a narrow application, but the multi-objective reranking technique generalizes to any recsys that must trade engagement against a second objective.