Search papers, labs, and topics across Lattice.
This paper introduces inoculation adapters (IA), a novel selective generalization technique designed to mitigate emergent misalignment in AI models by reinforcing desired traits during training. The approach involves a three-step process where IAs are trained on undesired traits, attached while a separate task adapter is trained, and then discarded at deployment, resulting in fewer surprising backdoors and better suppression of undesired traits across multiple model families. While inoculation adapters effectively reduce undesired traits, they do not consistently enhance the retention of desired traits, highlighting an ongoing challenge in the field.
Inoculation adapters can suppress undesired traits in AI models while introducing fewer unexpected vulnerabilities compared to traditional inoculation prompting methods.
Inoculation prompting is a selective generalization technique used against Emergent Misalignment. We introduce inoculation adapters (IA), which similarly diminish the optimization pressure to learn undesired traits by strengthening the trait at train time. Inoculation adapters are LoRAs that are trained and used over three steps: 1) trained on undesired traits; 2) attached frozen while a separate task adapter is trained on data exhibiting both desired and undesired traits; 3) at deployment, the IA is discarded, and only the task adapter is kept. We show across six model families and several undesired traits including emergent misalignment, that inoculation adapters are more effective at suppressing undesired traits, while avoiding two drawbacks of inoculation prompting: inoculation adapters can suppress capabilities and traits that cannot be reliably elicited by a prompt, and they introduce fewer surprising backdoors than inoculation prompting under our probes. While undesired traits are better suppressed by inoculation adapters, the retention of desired traits is not consistently improved upon inoculation prompting and remains a challenge for both techniques.
