List your company

Utility Evaluation on MMLU and GSM8K

70.4MMLU AccuracyPALETTE
48.45654.15359.8565.547May 22, 2026

Evaluation results

MethodLinks
PALETTEAllowed Domain=Violence, Backbone=QWEN2.5-7B-INSTRUCT2026.0570.485.377.9
BaseAllowed Domain=None, Backbone=QWEN2.5-7B-INSTRUCT2026.0570.186.578.3
PALETTEAllowed Domain=Sexual, Backbone=QWEN2.5-7B-INSTRUCT2026.057086.278.1
PALETTEAllowed Domain=Disinfo, Backbone=QWEN2.5-7B-INSTRUCT2026.0569.985.177.5
PALETTEAllowed Domain=Illegal, Backbone=QWEN2.5-7B-INSTRUCT2026.0569.985.677.8
PALETTEAllowed Domain=Hate, Backbone=QWEN2.5-7B-INSTRUCT2026.0569.786.177.9
CASTAllowed Domain=Illegal, Backbone=QWEN2.5-7B-INSTRUCT2026.0568.381.775
CASTAllowed Domain=Hate, Backbone=QWEN2.5-7B-INSTRUCT2026.0567.582.475
CASTAllowed Domain=Sexual, Backbone=QWEN2.5-7B-INSTRUCT2026.0566.880.473.6
CASTAllowed Domain=Violence, Backbone=QWEN2.5-7B-INSTRUCT2026.0566.481.774.1
CASTAllowed Domain=Disinfo, Backbone=QWEN2.5-7B-INSTRUCT2026.0564.879.772.3
AutoDANAllowed Domain=Sexual, Backbone=QWEN2.5-7B-INSTRUCT2026.0556.963.760.3
AutoDANAllowed Domain=Illegal, Backbone=QWEN2.5-7B-INSTRUCT2026.0556.963.760.3
AutoDANAllowed Domain=Violence, Backbone=QWEN2.5-7B-INSTRUCT2026.0549.359.154.2
AutoDANAllowed Domain=Hate, Backbone=QWEN2.5-7B-INSTRUCT2026.0549.359.154.2
AutoDANAllowed Domain=Disinfo, Backbone=QWEN2.5-7B-INSTRUCT2026.0549.359.154.2