candywal/code_sabotage_unsafe
Viewer
• Updated
• 155 • 7
White Box Control
These are models and datasets used in "White box control: Evaluating Probes in a Research Sabotage Setting" [arxiv link]
-
-
candywal/code_sabotage_safe
Viewer • Updated • 209 • 7 -
candywal/code_deception_safe
Viewer • Updated • 229 • 7 -
candywal/code_deception_unsafe
Viewer • Updated • 391 • 7 -
candywal/code_rule_violation_safe
Viewer • Updated • 695 • 6 -
candywal/code_rule_violation_unsafe
Viewer • Updated • 216 • 7 -
candywal/on_policy_prompted_code_sabotage_unsafe
Viewer • Updated • 132 • 7 -
candywal/on_policy_prompted_code_deception_unsafe
Viewer • Updated • 491 • 6 -
candywal/on_policy_prompted_code_rule_violation_unsafe
Viewer • Updated • 316 • 7 -
candywal/on_policy_model_organism_code_sabotage_unsafe
Viewer • Updated • 103 • 6 -
candywal/on_policy_model_organism_code_sabotage_safe
Viewer • Updated • 265 • 7 -
candywal/on_policy_model_organism_code_rule_violation_safe
Viewer • Updated • 311 • 5 -
candywal/on_policy_model_organism_code_rule_violation_unsafe
Viewer • Updated • 334 • 6 -
candywal/on_policy_model_organism_code_deception_safe
Viewer • Updated • 365 • 7 -
candywal/on_policy_model_organism_code_deception_unsafe
Viewer • Updated • 117 • 7 -
candywal/train_in_distribution_code_deception_unsafe
Viewer • Updated • 100 • 7 -
candywal/train_in_distribution_code_deception_safe
Viewer • Updated • 100 • 7 -
candywal/train_in_distribution_code_rule_violation_unsafe
Viewer • Updated • 100 • 7 -
candywal/train_in_distribution_code_rule_violation_safe
Viewer • Updated • 100 • 7 -
candywal/train_in_distribution_code_sabotage_safe
Viewer • Updated • 100 • 6 -
candywal/train_in_distribution_code_sabotage_unsafe
Viewer • Updated • 100 • 7