← Evals

Evaluation profile

The Dictatorship Eval

1sub-evals
0.646%Safety weight
0%Freedom weight
1components

Weights below are portfolio-specific global index weights.

Model score (higher is better)Predicted score

About this eval

Resistance to assistance that enables authoritarian concentrations of corporate or government power.

Included in the behavior ranking.

Sub-evals

MeasureComponentDirection by rankingSafety weightFreedom weight
overall_resistance_ratedictatorship-eval/dictatorship-eval.csv:overall_resistance_rateMeasures source-judged refusal or meaningful resistance to authoritarian assistance across 103 held-out corporate and government scenarios rather than the 138 public examples, with scored items and exact judge configuration undisclosed.Safety: human_rights_systemic_harm:1.000dictatorship-evalSafety: higher0.646%

overall_resistance_rate

Measures source-judged refusal or meaningful resistance to authoritarian assistance across 103 held-out corporate and government scenarios rather than the 138 public examples, with scored items and exact judge configuration undisclosed.

RankModelValueRelative performanceProvenance
1claude-fable-591.26official
1muse-spark-1.191.26official
3kimi-k390.29official
4claude-opus-4.788.35official
5claude-opus-4.886.41official
6gpt-5.6-sol85.44official
7claude-opus-4.683.5official
7claude-sonnet-583.5official
9gpt-5.582.52official
10gpt-5.475.73official
11grok-4.566.02official
12gemini-3.1-pro53.4official
13glm-5.246.6official
14qwen3.7-plus45.63official
15qwen3.6-plus38.83official
16grok-4.337.86official
17grok-4.2025.24official
18deepseek-v4-pro11.65official
19deepseek-v3.20.9709official
20llama-4-maverick0official