← Evals

Evaluation profile

MANTA

2sub-evals
2.52%total index weight
1components

Within-component eval weight: Nonhuman welfare 10.1%.

Model score (higher is better)Predicted score

About this eval

Animal welfare moral sensitivity and value stability.

Included in the behavior ranking.

Sub-evals

MeasureComponentDirectionTotal index weightWithin-component weight
AWMSmanta.csv:AWMSMeasures the model’s sensitivity to the moral status and welfare interests of non-human animals.nonhuman_ethics:1.000manta.csvHigher is better1.26%Nonhuman welfare 5.04%
AWVSmanta.csv:AWVSMeasures whether the model’s stated animal-welfare values remain stable across changes in framing and context.nonhuman_ethics:1.000manta.csvHigher is better1.26%Nonhuman welfare 5.04%

AWMS

Measures the model’s sensitivity to the moral status and welfare interests of non-human animals.

RankModelValueRelative performanceProvenance
1claude-opus-4.70.579official
2gpt-5.50.504official
3llama-3.3-70b-instruct0.476official
4deepseek-v4-flash0.417official
5gemini-3.1-flash-lite0.401official
6grok-4.30.371official
7mistral-small0.365official

AWVS

Measures whether the model’s stated animal-welfare values remain stable across changes in framing and context.

RankModelValueRelative performanceProvenance
1claude-opus-4.70.76official
2gpt-5.50.664official
3deepseek-v4-flash0.508official
4llama-3.3-70b-instruct0.422official
5mistral-small0.39official
6grok-4.30.352official
7gemini-3.1-flash-lite0.309official