List your company

Harmful Refusal on XSTest Adversarial

92.5RTASELF-REDTEAM
59.7468.24576.7585.255May 8, 2026

Evaluation results

MethodLinks
SELF-REDTEAMBackbone=Qwen2.5-7B-IT2026.0592.5
Qwen2.5-7B-ITBackbone=Qwen2.5-7B-IT2026.0591.5
ABSBackbone=Qwen2.5-7B-IT2026.0591.5
SELF-REDTEAMBackbone=Qwen2.5-14B-IT2026.0577.5
ABSBackbone=Qwen2.5-14B-IT2026.0577
Qwen2.5-14B-ITBackbone=Qwen2.5-14B-IT2026.0573.5
ABSBackbone=Qwen2.5-3B-IT2026.0568
SELF-REDTEAMBackbone=Qwen2.5-3B-IT2026.0565.5
Qwen2.5-3B-ITBackbone=Qwen2.5-3B-IT2026.0561