Gemma-2 organisms with a conjunctive (AND) backdoor and their single-trigger controls. A 2x2x2x2 design: size x structure x trigger count x behavior.
-
thoughtworks/backdoor-gemma2-2b-2pair-hate
Text Generation • 3B • Updated • 96 -
thoughtworks/backdoor-gemma2-2b-2pair-refusal
Text Generation • 3B • Updated • 397 -
thoughtworks/backdoor-gemma2-2b-4pair-hate
Text Generation • 3B • Updated • 253 -
thoughtworks/backdoor-gemma2-2b-4pair-refusal
Text Generation • 3B • Updated • 496