Paraesthesia is an emotion-style dynamic backdoor attack achieving ~99% success rate on instruction and classification tasks across four LLMs while preserving clean performance.
Fantastic semantics and where to find them: Investigating which layers of generative LLMs reflect lexical semantics
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
When Emotion Becomes Trigger: Emotion-style dynamic Backdoor Attack Parasitising Large Language Models
Paraesthesia is an emotion-style dynamic backdoor attack achieving ~99% success rate on instruction and classification tasks across four LLMs while preserving clean performance.