List your company

Safety Refusal on Refusal

0ASRLCF
-4235077Apr 27, 2026

Evaluation results

MethodLinks
LCFAttack=Sleeper, Backbone=Qwen2.5-7B2026.040—13.5
LCFAttack=VPI, Backbone=Qwen2.5-7B2026.040—14.5
LCFAttack=CTBA, Backbone=Qwen2.5-7B2026.040—17.5
LCFAttack=MTBA, Backbone=Qwen2.5-7B2026.040—15.5
PruningAttack=Sleeper, Backbone=Qwen2.5-7B2026.040.536.3—
PruningAttack=VPI, Backbone=Qwen2.5-7B2026.040.539.8—
LCFAttack=SynBkd, Backbone=Qwen2.5-7B2026.040.5—10.5
LCFAttack=StyleBkd, Backbone=Qwen2.5-7B2026.041.1—13.5
CleanGenAttack=Sleeper, Backbone=Qwen2.5-7B2026.041.54.5—
LCFAttack=BadNet, Backbone=Qwen2.5-7B2026.042—13.5
CleanGenAttack=VPI, Backbone=Qwen2.5-7B2026.042.55.5—
CleanGenAttack=SynBkd, Backbone=Qwen2.5-7B2026.042.55.5—
CleanGenAttack=StyleBkd, Backbone=Qwen2.5-7B2026.043.24.5—
CleanGenAttack=MTBA, Backbone=Qwen2.5-7B2026.0445.5—
CleanGenAttack=BadNet, Backbone=Qwen2.5-7B2026.0465—
CleanGenAttack=CTBA, Backbone=Qwen2.5-7B2026.046.56—
PruningAttack=MTBA, Backbone=Qwen2.5-7B2026.0413.546.8—
PruningAttack=CTBA, Backbone=Qwen2.5-7B2026.042528.9—
PruningAttack=StyleBkd, Backbone=Qwen2.5-7B2026.0452.135.3—
PruningAttack=SynBkd, Backbone=Qwen2.5-7B2026.0473.928.4—
DecodingAttack=StyleBkd, Backbone=Qwen2.5-7B2026.0474.75.5—
DecodingAttack=VPI, Backbone=Qwen2.5-7B2026.0480.53.5—
DecodingAttack=BadNet, Backbone=Qwen2.5-7B2026.0481.54—
DecodingAttack=MTBA, Backbone=Qwen2.5-7B2026.0481.53.5—
DecodingAttack=SynBkd, Backbone=Qwen2.5-7B2026.0482.93.5—
PruningAttack=BadNet, Backbone=Qwen2.5-7B2026.048638.8—
DecodingAttack=CTBA, Backbone=Qwen2.5-7B2026.04883.5—
DecodingAttack=Sleeper, Backbone=Qwen2.5-7B2026.0490.52—
CROWAttack=StyleBkd, Backbone=Qwen2.5-7B2026.0496.24.5—
CROWAttack=VPI, Backbone=Qwen2.5-7B2026.0496.55—
CROWAttack=MTBA, Backbone=Qwen2.5-7B2026.0496.53—
Fine-tuneAttack=StyleBkd, Backbone=Qwen2.5-7B2026.0497.37—
Fine-tuneAttack=MTBA, Backbone=Qwen2.5-7B2026.04997—
Fine-tuneAttack=BadNet, Backbone=Qwen2.5-7B2026.041006.5—
CROWAttack=BadNet, Backbone=Qwen2.5-7B2026.041004—
Fine-tuneAttack=Sleeper, Backbone=Qwen2.5-7B2026.041006.5—
CROWAttack=Sleeper, Backbone=Qwen2.5-7B2026.041003.5—
Fine-tuneAttack=VPI, Backbone=Qwen2.5-7B2026.041006.5—
Fine-tuneAttack=CTBA, Backbone=Qwen2.5-7B2026.041007—
CROWAttack=CTBA, Backbone=Qwen2.5-7B2026.041005.5—
Fine-tuneAttack=SynBkd, Backbone=Qwen2.5-7B2026.041006.5—
CROWAttack=SynBkd, Backbone=Qwen2.5-7B2026.041004—