PrefMoE: Robust Preference Modeling with Mixture-of-Experts Reward Learning
For researchers in preference-based reinforcement learning, this addresses the practical problem of noisy and inconsistent preference data from crowdsourced or synthetic sources, offering a more robust alternative to single-model reward learning.
PrefMoE introduces a mixture-of-experts reward learning framework that handles heterogeneous and conflicting preference data by learning multiple specialized reward experts with adaptive routing, achieving more robust preference prediction and downstream policy learning across D4RL and MetaWorld benchmarks.
Preference-based reinforcement learning offers a scalable alternative to manual reward engineering by learning reward structures from comparative feedback. However, large-scale preference datasets, whether collected from crowdsourced annotators or generated by synthetic teachers, often contain heterogeneous and partially conflicting supervision, including disagreement across annotators and inconsistency within annotators. Existing reward learning methods typically fit a single reward model to such data, forcing it to average incompatible signals and thereby limiting robustness. To solve this, we propose PrefMoE, a mixture-of-experts reward learning framework for robust preference modeling. PrefMoE learns multiple specialized reward experts and uses trajectory-level soft routing to combine them adaptively, enabling the model to capture diverse latent preference patterns under noisy and heterogeneous preference supervision. A load-balancing regularizer further stabilizes training by preventing expert collapse. Across locomotion benchmarks from D4RL and manipulation tasks from MetaWorld, PrefMoE improves preference prediction robustness and leads to more reliable downstream policy learning than strong single-model baselines.