AI Safety Gridworlds

AI Safety Gridworlds is a benchmark that contains a set of tasks designed to evaluate the safety and robustness of AI systems, particularly in the context of reward hacking and specification gaming in reinforcement learning scenarios.

Papers using AI Safety Gridworlds (2)

Reward Hacking in Language Model Agents: Revisiting AI Safety Gridworlds2026Discovering Agentic Safety Specifications From 1-bit Danger Signals2026