UMM-RM: An Upcycle-and-Merge MoE Reward Model for Mitigating Reward Hacking

arXiv:2512.00724 · cs.LG, cs.IR · Submitted 2025-11-30 · Read on arXiv

cs.LG, cs.IR

Submitted: 2025-11-30

Updated: 2026-10-11

Terminology

• Reward Hacking — 3× in this paper · explained in 8 episodes
• Alignment — 2× in this paper · explained in 1 episode
• Upcycle-and-Merge MoE Reward Model — 2× in this paper
• Accuracy — 1× in this paper · explained in 1 episode
• Distribution Shift — 1× in this paper · explained in 7 episodes
• Mixture-of-Experts (MoE) — 1× in this paper · explained in 7 episodes
• Reinforcement Learning — 1× in this paper · explained in 4 episodes
• Robustness — 1× in this paper · explained in 5 episodes

Related papers

• Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
• AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
• Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
• Convergence issues in Relational Concept Analysis based on AOC-posets
• Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
• Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks