LLMs that can decrypt common ciphers show safety failures split across two dimensions, refusing too much or generating unsafe output, and current defenses fix one side while breaking the other.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Should LLM Safety Be More Than Refusing Harmful Instructions?
LLMs that can decrypt common ciphers show safety failures split across two dimensions, refusing too much or generating unsafe output, and current defenses fix one side while breaking the other.