Breaking Free from Hand-Crafted Rewards: A Genetic Programming Framework for End-Goal-Driven Reinforcement Learning

Created by W.Langdon from gp-bibliography.bib Revision:1.9159

@InProceedings{Kumar:2026:CEC,

• author = "Kamalesh Kumar and Jean-Alexis Delamer and James Alexander Hughes",
• title = "Breaking Free from Hand-Crafted Rewards: A Genetic Programming Framework for End-Goal-Driven Reinforcement Learning",
• booktitle = "IEEE CEC 2026",
• year = "2026",
• editor = "Nelishia Pillay and Yew-Soon Ong and Carlos Coello and Juergen Branke and Simon Lucas and Chuan-Kang Ting and Bing Xue",
• pages = "cec pap134",
• address = "Maastricht, the Netherlands",
• month = "21-26 " # jun,
• publisher = "IEEE",
• keywords = "genetic algorithms, genetic programming, End-Goal-Driven Fitness, Reinforcement Learning, Reward Function",
• URL = "https://wcci.klinkhamergroup.com/papers/by_date.html#cec_pap134",
• URL = "https://linklings.s3.amazonaws.com/organizations/WCCI/wcci2026/submissions/stype102/nZzyN-cec_pap134s2.pdf",
• size = "8 pages",
• abstract = "... We introduce RevolveRL, a novel approach leveraging Genetic Programming to discover effective reward functions that enhance performance and reduce training time. Our method employs task-specific fitness functions designed around the ultimate goals of the tasks, which guide the evolutionary search process. We evaluate our approach on both classic RL benchmarks and the more complex MuJoCo physics simulation tasks. Our experimental results demonstrate that automatically evolved reward functions consistently yield superior policies and accelerate the training process compared to baseline manually designed rewards.",

}

Genetic Programming entries for Kamalesh Kumar Jean-Alexis Delamer James Alexander Hughes

Citations