A trigger-agnostic defense that merges a backdoored LLM with a clean-data LoRA model and adds contradictory prompt evidence, reducing attack success while keeping most clean-task accuracy.
Atteq- nn: Attention-based qoe-aware evasive backdoor attacks
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Neutralizing Backdoors through Information Conflicts for Large Language Models
A trigger-agnostic defense that merges a backdoored LLM with a clean-data LoRA model and adds contradictory prompt evidence, reducing attack success while keeping most clean-task accuracy.