Adversarial Reward Hacking
Fresh
Adversarial Reward Hacking Environment: Multi-vector benchmark for studying specification gaming, sycophancy, and proxy-true reward divergence in RLHF
- Type
- RL Env
- License
- mit
- Size
- v0.0.1
- Published
- Sep 2026
- Updated
- Sep 2026
Cite
Notes
Only stored in your browser.