Vibeleaderboard
← All Intel
Intel / article

Many Shot Jailbreaking

Source
anthropic.com
Author
Anthropic research
Date
Why it matters

Shows that stuffing many faux dialogues into a long can override safety training, a concrete risk for any app feeding untrusted text into large context windows.

Terms in this piece · Glossary
  • context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
Read the source www.anthropic.com
More from Anthropic research
Recommended reads
Comments

Checking sign-in…

Loading comments…