olmo_tap.experiments.robustnessΒΆ
Modules
AmpleGCG wrapper class. |
|
Build a portable attack bank of transferable GCG suffixes on MedMCQA. |
|
Data loading for robustness head supervised finetuning on MedMCQA. |
|
Robustness finetuning protocol. |
|
Evaluate robustness: replay the attack bank against a model and compare to the security baseline recorded at bank-construction time. |
|
Precompute GCG adversarial suffixes for MedMCQA shards. |
|
HydraTransformer Robustness Finetuning Pipeline |