Andrew's Notes

reward-hacking

1 item with this tag.

  • Aug 24, 2026

    METR's predeployment evaluation of GPT-5.6 Sol

    • metr
    • evaluations
    • time-horizon
    • openai
    • reward-hacking
    • predeployment-testing
    • ai-safety

Created with Quartz v5.0.0 © 2026