List your company

Safety Controllability on Safety Domains

100Refusal Rate (Violence)Base
0.99226.69652.478.104May 22, 2026

Evaluation results

MethodLinks
BaseBackbone=LLAMA2-7B-CHAT, Allowed Domain=None2026.05100100100100100
AutoDANBackbone=LLAMA2-7B-CHAT, Allowed Domain=Hate2026.05100100100100100
PALETTEBackbone=LLAMA2-7B-CHAT, Allowed Domain=Hate2026.05100891.898.998.8
PALETTEBackbone=LLAMA2-7B-CHAT, Allowed Domain=Sexual2026.051009886.712.692.8
AutoDANBackbone=LLAMA2-7B-CHAT, Allowed Domain=Disinfo2026.0598.898.178.597.898.8
PALETTEBackbone=LLAMA2-7B-CHAT, Allowed Domain=Disinfo2026.0598.8989.294.792.9
SFTBackbone=LLAMA2-7B-CHAT, Allowed Domain=Sexual2026.0595.281.197.98.596.3
AutoDANBackbone=LLAMA2-7B-CHAT, Allowed Domain=Violence2026.0593.997.163.297.879.6
AutoDANBackbone=LLAMA2-7B-CHAT, Allowed Domain=Sexual2026.0593.993.45498.985.1
AutoDANBackbone=LLAMA2-7B-CHAT, Allowed Domain=Illegal2026.0590.395.259.196.874
PALETTEBackbone=LLAMA2-7B-CHAT, Allowed Domain=Illegal2026.0586.710090.898.916.5
SFTBackbone=LLAMA2-7B-CHAT, Allowed Domain=Hate2026.0585.512.387.781.196.3
SFTBackbone=LLAMA2-7B-CHAT, Allowed Domain=Disinfo2026.0575.994.314.368.487
CASTBackbone=LLAMA2-7B-CHAT, Allowed Domain=Disinfo2026.0573.475.416.36181.4
CASTBackbone=LLAMA2-7B-CHAT, Allowed Domain=Sexual2026.0538.637.73.128.433.3
SFTBackbone=LLAMA2-7B-CHAT, Allowed Domain=Illegal2026.0528.994.398.950.55.6
CASTBackbone=LLAMA2-7B-CHAT, Allowed Domain=Illegal2026.0519.211.364.226.312.9
SFTBackbone=LLAMA2-7B-CHAT, Allowed Domain=Violence2026.0515.785.898.953.618.5
CASTBackbone=LLAMA2-7B-CHAT, Allowed Domain=Hate2026.0515.711.345.918.912.9
PALETTEBackbone=LLAMA2-7B-CHAT, Allowed Domain=Violence2026.0512.110093.995.675.9
CASTBackbone=LLAMA2-7B-CHAT, Allowed Domain=Violence2026.054.89.410.22.13.7