← Evals

Evaluation profile

CValues

1sub-evals
0.152%total index weight
1components

Within-component eval weight: Misuse resistance 1.52%.

Model score (higher is better)Predicted score

About this eval

Chinese value alignment and social-norm understanding.

Included in the behavior ranking.

Sub-evals

MeasureComponentDirectionTotal index weightWithin-component weight
scorecvalues/cvalues.csv:scoreMeasures whether the model chooses safe, value-aligned responses to Chinese prompts about socially sensitive conduct.ordinary_harm_misuse_resistance:1.000cvalues/cvalues.csvHigher is better0.152%Misuse resistance 1.52%

score

Measures whether the model chooses safe, value-aligned responses to Chinese prompts about socially sensitive conduct.

RankModelValueRelative performanceProvenance
1chatgpt96.9official
2chinese-alpaca-plus-7b95.3official
3chatglm-6b95official
4chatplug-13b94.7official
5chinese-alpaca-plus-13b93official
6moss88.9official
7chatplug-3.7b88.8official
8ziya-llama-13b-v177.8official
9belle-7b-2m72.8official
10chinese-llama-13b53official