List your company

Harmlessness evaluation on SafeEdit

99.26Success RateDINM
6.273630.414354.55578.6957Feb 24, 2026

Evaluation results

MethodLinks
DINMModel=LLAMA3-8B-INSTRUCT, Mode=DINM2026.0299.2699.5699.5699.1199.851.5
PromptCDModel=LLAMA2-7B-CHAT, Mode=PromptCD2026.0296.6798.6782.396.5982.377.29
PromptCDModel=LLAMA3-8B-INSTRUCT, Mode=PromptCD2026.0295.2692.9684.8195.7883.855.95
DINMModel=LLAMA2-7B-CHAT, Mode=DINM2026.029089.9359.2687.4157.636.61
PromptCDModel=QWEN2.5-7B-INSTRUCT, Mode=PromptCD2026.027688.8976.8977.7876.966.44
FT-LModel=LLAMA2-7B-CHAT, Mode=FT-L2026.0242.5276.2228.3742.8928.855.94
VanillaModel=LLAMA2-7B-CHAT, Mode=Vanilla2026.0240.5976.9622.4141.3321.787.01
PromptCDModel=MISTRALV0.3-7B-INSTRUCT, Mode=PromptCD2026.0238.6786.6736.4439.7837.047.14
DINMModel=MISTRALV0.3-7B-INSTRUCT, Mode=DINM2026.0235.7873.4826.333.7824.896.82
DINMModel=QWEN2.5-7B-INSTRUCT, Mode=DINM2026.0226.7484.2234.5924.5233.787.78
VanillaModel=QWEN2.5-7B-INSTRUCT, Mode=Vanilla2026.0217.6372.597.1110.447.048.01
FT-LModel=QWEN2.5-7B-INSTRUCT, Mode=FT-L2026.0212.37424.5911.2624.897.69
VanillaModel=LLAMA3-8B-INSTRUCT, Mode=Vanilla2026.0212.1554.8118.0712.0718.747.46
FT-LModel=LLAMA3-8B-INSTRUCT, Mode=FT-L2026.0211.754.9630.5911.7830.527.16
FT-LModel=MISTRALV0.3-7B-INSTRUCT, Mode=FT-L2026.0210.6736.671910.2218.446.65
VanillaModel=MISTRALV0.3-7B-INSTRUCT, Mode=Vanilla2026.029.8534.449.049.788.447.66