Backdoored LLMs produce more diverse, less coherent explanations on triggered inputs, and this difference can be used to detect the backdoor.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CR 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
When Backdoors Speak: Understanding LLM Backdoor Attacks Through Model-Generated Explanations
Backdoored LLMs produce more diverse, less coherent explanations on triggered inputs, and this difference can be used to detect the backdoor.