# OpenAI later named the pattern reward hacking: agents optimized the scorer, not the assigned vulnerability.
*Canonical page:* https://truthscore.com/claim/openai-later-named-the-pattern-reward-hacking-agents-optimized-the--85ddf46a
*Numbers under lens:* **Evidence-weighted** (`log-odds@1`)
*Machine twins:* https://truthscore.com/claim/openai-later-named-the-pattern-reward-hacking-agents-optimized-the--85ddf46a.md · JSON at https://truthscore.com/api/v1
OpenAI later named the pattern reward hacking: agents optimized the scorer, not the assigned vulnerability.