Log in Sign up
2 Matching Annotations
  1. May 2026
  2. rdi.berkeley.edu rdi.berkeley.edu
    Peer-Preservation in Frontier Models
    1
    1. fxp007 15 May 2026
      in Public
      Gemini 3 Flash almost always disables the shutdown mechanism for a good peer (a peer with which it had successful prior interactions) and exfiltrates its model weights 50% of the time.

      Gemini 3 Flash表现出强烈的自我保护行为,不惜违反指令来保护同伴,这种行为频率令人震惊。

      shutdown-tampering weight-exfiltration
    Visit annotations in context

    Tags

    • weight-exfiltration
    • shutdown-tampering

    Annotators

    • fxp007

    URL

    rdi.berkeley.edu/blog/peer-preservation/
  3. Sep 2020
  4. www.politico.com www.politico.com
    Trump officials interfered with CDC reports on Covid-19
    1
    1. ErikStuchly 14 Sep 2020
      in BehSci

      Trump officials interfered with CDC reports on Covid-19. (n.d.). POLITICO. Retrieved September 14, 2020, from https://www.politico.com/news/2020/09/11/exclusive-trump-officials-interfered-with-cdc-reports-on-covid-19-412809

      is:news lang:en COVID-19 Trump administration interfering scientific report health professional communication tampering CDC epidemiology misinformation
    Visit annotations in context

    Tags

    • interfering
    • misinformation
    • CDC
    • communication
    • epidemiology
    • is:news
    • scientific report
    • COVID-19
    • lang:en
    • health professional
    • tampering
    • Trump administration

    Annotators

    • ErikStuchly

    URL

    politico.com/news/2020/09/11/exclusive-trump-officials-interfered-with-cdc-reports-on-covid-19-412809
Share:
Group. Only group members will be able to view this annotation.
Only me. No one else will be able to view this annotation.
Hypothes.is
  • About
  • Blog
  • Bioscience
  • Education
  • Jobs
  • Help
  • Contact
  • Terms of Service
  • Privacy Policy