Transplanting, inverting, and preventing a misalignment persona: method-conditional emergent misalignment in Qwen2.5 · 相关技术
相关技术
SFTLoRAPEFTACTAbstractADCANNARROWATEAbstract Syntax TreeAction UnitsAdaptive Batch ScalingAdaptive Subset Selection AttackAgents' Last ExamAirborne Laser Scanning (ALS)Attribute Prefix AdapterAugmented Latent IntrinsicsBlurry Window AttentionCAPCASTCERCURECalibrated Entropy ScoreCamouflaged object detectionCardiology Interface TerminologyCartridges at Scale (CAS)Condition Dropout (ConD)Cryo-electron microscopyCyber Threat IntelligenceDCASDELDIRDIRECTDIffusion Reconstruction ErrorDNA-encoded librariesDOMDensity-Aware TranslationDiTDynamic Infilling AnchorsECTELSpretraining