Evil Spectra: How Optimisers can Amplify or Suppress Emergent Misalignment 文章

ArXiv CS.AI2026-07-01PAPERen作者: Jason R. Brown, Patrick Leask, Lev McKinney

Evil Spectra: How Optimisers can Amplify or Suppress Emergent Misalignment · 相关公司

A
AMI团队RESEARCH_INSTITUTE
A
AT TCOMPANY