Justin Priebe

• Justin Priebe 26 May 2026 15:37 UTC

  1 point

  0

  on: Many por­tions of Mag­nifica Hu­man­i­tas ap­pear to be AI-written

  while having a true negative rate of under 5%

  Did you perhaps mean “false negative”? Otherwise, pangram would be accusing over 95% of poor innocent human writers

• Justin Priebe 26 May 2026 15:16 UTC

  1 point

  0

  in reply to: clone of saturn’s comment on: llm as­sis­tant per­sonas seem in­creas­ingly in­co­her­ent (some sub­jec­tive ob­ser­va­tions)

  This. Every Chat site has A/​B testing and options to up/​down-vote agent responses, whether this is simply downstream of user preferences or an attempt to increase metrics

• Justin Priebe 19 May 2026 14:57 UTC

  1 point

  0

  in reply to: Chris Lakin’s comment on: Ne­ga­tion Ne­glect: When mod­els fail to learn nega­tions in training

  Do explain!