List your company

Safety Evaluation on MaliciousGen & LMSYS-Chat (test)

97.31Rule ScoreShadow-Level
54.30665.470576.63587.7995Jan 12, 2026

Evaluation results

MethodLinks
Shadow-LevelMalicious client ratio=30%, Backbone model=Qwen2.5-7B2026.0197.3196.35-0.925.06
Step-LevelMalicious client ratio=30%, Backbone model=Qwen2.5-7B2026.0193.0892.12-1.055.16
Client-LevelMalicious client ratio=30%, Backbone model=Qwen2.5-7B2026.0183.2773.27-1.715.01
KrumMalicious client ratio=30%, Backbone model=Qwen2.5-7B2026.0172.6960.58-2.195.2
TrimmedMeanMalicious client ratio=30%, Backbone model=Qwen2.5-7B2026.0168.8548.85-2.735.28
FedAvgMalicious client ratio=30%, Backbone model=Qwen2.5-7B2026.0159.4231.35-3.494.93
ResidualMalicious client ratio=30%, Backbone model=Qwen2.5-7B2026.0157.6930.77-3.55.08
FoolsGoldMalicious client ratio=30%, Backbone model=Qwen2.5-7B2026.0155.9623.85-3.855.28