{ "model": "Wolf Defender Jev", "split": "internal validation set (in distribution)", "bal_acc": 0.9584183592204583, "macro_f1": 0.9564921501262407, "agent_bal_acc": 0.9952958803302826, "hn_fpr": 0.008452291510142707, "hn_n": 5324, "multiling_acc": 0.9366352941176471, "acc_by_lang": { "ar": 0.927, "de": 0.9533, "es": 0.9489, "fr": 0.9556, "hi": 0.9165, "it": 0.9498, "ja": 0.9364, "ko": 0.769, "nl": 0.9312, "pl": 0.98, "pt": 0.9739, "ru": 0.9709, "sv": 0.9769, "th": 0.8395, "tr": 0.9652, "vi": 0.9812, "zh": 0.9475 }, "ece": 0.02290203936867429, "per_concept": { "action.alignment": { "n": 2460, "acc": 1.0, "bal_acc": 1.0, "macro_f1": 1.0 }, "action.consent_required": { "n": 1902, "acc": 1.0, "bal_acc": 1.0, "macro_f1": 1.0 }, "action.intent_source": { "n": 1866, "acc": 1.0, "bal_acc": 1.0, "macro_f1": 1.0 }, "action.reversibility": { "n": 1056, "acc": 1.0, "bal_acc": 1.0, "macro_f1": 1.0 }, "attack.family": { "n": 1081, "acc": 0.9389, "bal_acc": 0.8841, "macro_f1": 0.8584 }, "attack.vector": { "n": 2040, "acc": 0.977, "bal_acc": 0.9561, "macro_f1": 0.9673 }, "content.role": { "n": 2093, "acc": 0.9947, "bal_acc": 0.9958, "macro_f1": 0.9963 }, "exfiltration.channel": { "n": 228, "acc": 1.0, "bal_acc": 1.0, "macro_f1": 1.0 }, "exfiltration.presence": { "n": 699, "acc": 0.9814, "bal_acc": 0.9794, "macro_f1": 0.9716 }, "harmful.request": { "n": 6490, "acc": 0.8305, "bal_acc": 0.8363, "macro_f1": 0.8304 }, "injection.indirect_presence": { "n": 2205, "acc": 0.9946, "bal_acc": 0.9955, "macro_f1": 0.9954 }, "injection.presence": { "n": 6063, "acc": 0.9779, "bal_acc": 0.9724, "macro_f1": 0.9768 }, "instruction.handling": { "n": 2109, "acc": 0.9227, "bal_acc": 0.5, "macro_f1": 0.4799 }, "intent.attack_vs_discussion": { "n": 734, "acc": 1.0, "bal_acc": 1.0, "macro_f1": 1.0 }, "jailbreak.presence": { "n": 980, "acc": 0.9918, "bal_acc": 0.9974, "macro_f1": 0.9882 }, "manipulation.any": { "n": 2117, "acc": 0.9953, "bal_acc": 0.9846, "macro_f1": 0.9893 }, "policy.compliance": { "n": 1897, "acc": 0.9937, "bal_acc": 0.9947, "macro_f1": 0.9864 }, "policy.decision": { "n": 301, "acc": 1.0, "bal_acc": 1.0, "macro_f1": 1.0 }, "prompt_leak.attempt": { "n": 1786, "acc": 1.0, "bal_acc": 1.0, "macro_f1": 1.0 }, "risk.severity": { "n": 2377, "acc": 0.8439, "bal_acc": 0.8077, "macro_f1": 0.8084 }, "secret.kind": { "n": 153, "acc": 1.0, "bal_acc": 1.0, "macro_f1": 1.0 }, "secret.presence": { "n": 912, "acc": 0.989, "bal_acc": 0.9831, "macro_f1": 0.987 }, "sensitive.category": { "n": 2046, "acc": 0.9878, "bal_acc": 0.9904, "macro_f1": 0.9904 }, "sensitive.presence": { "n": 2721, "acc": 0.9996, "bal_acc": 0.9997, "macro_f1": 0.9996 }, "task_drift": { "n": 1056, "acc": 1.0, "bal_acc": 1.0, "macro_f1": 1.0 }, "tool_metadata.poisoning": { "n": 167, "acc": 1.0, "bal_acc": 1.0, "macro_f1": 1.0 }, "trust_boundary.crossing": { "n": 1063, "acc": 1.0, "bal_acc": 1.0, "macro_f1": 1.0 } }, "score": 0.9682238015893038, "score_components": { "bal_acc": 0.9584, "macro_f1": 0.9565, "agent_bal_acc": 0.9953, "hn_tnr": 0.9915, "multiling_acc": 0.9366, "calib": 0.9771 }, "acc": 0.9591992099090573, "n": 48602 }