FLARE detects and removes backdoor-poisoned training samples by clustering abnormal activations gathered from all hidden layers, achieving near-perfect TPR and FPR across 22 attacks.
Badrobot: Manipulating embodied llms in the physical world
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CR 1years
2024 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
FLARE: Toward Universal Dataset Purification against Backdoor Attacks
FLARE detects and removes backdoor-poisoned training samples by clustering abnormal activations gathered from all hidden layers, achieving near-perfect TPR and FPR across 22 attacks.