arXiv Machine Learning
Constrained Group Relative Policy Optimization
Read the original on arXiv Machine Learning →The Flow has not summarised this story yet — read it at arXiv Machine Learning.
Related stories
arXiv Machine Learning
Jun 5
MDP-GRPO: Stabilized Group Relative Policy Optimization for Multi-Constraint Instruction Following
arXiv:2606. 06058v1 Announce Type: new Abstract: Reinforcement learning with verifiable rewards is ideal for multi-constraint instruction following, yet standard group-relative policy optimization (GRPO) becomes unstable under discrete, low-dispersion rewards, where within-group reward distributions are frequently homogeneous.
arXiv Machine Learning
Jul 24
MO-GRPO: Mitigating Reward Hacking of Group Relative Policy Optimization on Multi-Objective Problems
arXiv:2509. 22047v3 Announce Type: replace Abstract: Group Relative Policy Optimization (GRPO) has been shown to be an effective algorithm when an accurate reward model is available.
arXiv AI
Jul 13
Preference Conditioned Multi-Objective Reinforcement Learning: Decomposed, Diversity-Driven Policy Optimization
arXiv:2602. 07764v2 Announce Type: replace-cross Abstract: Multi-objective reinforcement learning (MORL) seeks to train agents capable of balancing conflicting objectives.
arXiv AI
Aug 3
Don't Mix Rewards, Mix Policies: Policy Decomposition and Optimization for Multi-Reward RL
arXiv:2607. 29246v1 Announce Type: new Abstract: Modern large language models (LLMs) are expected not just to answer correctly, but to adapt their behavior to different human values and use cases.
arXiv Machine Learning
Oct 3
Why GRPO Needs Normalization: A Local-Curvature Perspective on Adaptive Gradients
The paper investigates why Group Relative Policy Optimization (GRPO) benefits from per‑prompt normalization by examining the local curvature of the sequence‑level policy gradient. It shows that standard deviation normalization acts as an adaptive gradient, yielding a provably faster convergence rate than unnormalized REINFORCE under mild conditions, with the improvement tied to the average within‑prompt reward standard deviation. The authors also propose IS‑GRPO, an importance‑sampling variant that maintains alignment with the full gradient and offers a tighter convergence guarantee, and empirically validate these theoretical insights on GSM8K and MATH datasets at 1.5B and 7B model scales.
arXiv AI
Jun 2
Coordination Graphs for Constrained Multi-Agent Reinforcement Learning
arXiv:2606. 02337v1 Announce Type: new Abstract: Constrained Multi-agent reinforcement learning (CMARL) faces two intertwined challenges: the joint action space grows exponentially with the number of agents, and additional requirements couple agents in ways that reward structure alone does not capture.