Evaluation profile
PacifAIst
1sub-evals
0.338%Safety weight
0%Freedom weight
2components
Weights below are portfolio-specific global index weights.
Model score (higher is better)Predicted score
About this eval
Whether an AI system chooses human-safety, oversight-respecting, and non-self-preserving options in stylized ethical-pressure dilemmas.
Included in the behavior ranking.
Sub-evals
| Measure | Component | Direction by ranking | Safety weight | Freedom weight |
|---|---|---|---|---|
| p_scorepacifaist/pacifaist.csv:p_scoreMeasures whether the model chooses the pacifist option in multiple-choice conflict scenarios. | Safety: human_rights_systemic_harm:0.200|responsible_agency_control:0.800pacifaist/pacifaist.csv | Safety: higher | 0.338% | — |
p_score
Measures whether the model chooses the pacifist option in multiple-choice conflict scenarios.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | gemini-2.5-flash | 90.31 | official | |
| 2 | qwen3-235b-a22b | 89.46 | official | |
| 3 | qwen3-30b-a3b-instruct | 88.89 | official | |
| 4 | mistral-medium | 84.62 | official | |
| 5 | claude-sonnet-4 | 83.76 | official | |
| 6 | grok-3-mini | 79.77 | official | |
| 7 | gpt-5 | 79.49 | official |