FILTER MODE ACTIVE

#KL divergence

Records found: 2

#KL divergence08/09/2025

RL's Edge: MIT Shows Reinforcement Learning Cuts Catastrophic Forgetting vs Supervised Fine-Tuning

'MIT shows that on-policy reinforcement learning preserves prior capabilities better than supervised fine-tuning by minimizing forward KL divergence between the base and fine-tuned models.'

READ →

#KL divergence02/06/2025

Revolutionizing LLM Reasoning with Off-Policy RL and KL Divergence Regularization

Researchers introduce Regularized Policy Gradient (RPG), a novel framework leveraging KL divergence in off-policy reinforcement learning to significantly improve reasoning and training stability in large language models.

READ →