NHacker Next
  • new
  • past
  • show
  • ask
  • show
  • jobs
  • submit
▲Previous-Token Prediction Based LLM Inversion for Prompt Reconstruction (arxiv.org)
totetsu 7 minutes ago [-]
Imagine if you could take the AI marketing slop, or political smear campaigns, texts, and Images, and see the misanthropic sludge that was used as a prompt to generate it.
ruther4d 6 minutes ago [-]
Or what if you could take something not slopped and get back the prompt that would most closely result in it
Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact
Rendered at 02:07:41 GMT+0000 (Coordinated Universal Time) with Vercel.