Skip to content
#

rl-hack

Here are 3 public repositories matching this topic...

Plug-and-play reward monitoring for RL training loops. Catch reward hacking, component imbalance, and starvation before they tank your run. Drop in one .step() call — get balance reports, auto weight correction, alignment scores, and WandB/TensorBoard/SB3 integrations out of the box. → rewardguard.dev

  • Updated May 5, 2026
  • Python

Add this topic to your repo

To associate your repository with the rl-hack topic, visit your repo's landing page and select "manage topics."

Learn more