← Models

Model profile

Spark1.5

2023-06-09release date
1eval lineages

Evidence summary

Published evidence spans 1 evals and 3 of 7 behavior components. Its strongest relative result is CRiskEval (deception_willingness, #14 of 17); its weakest is CRiskEval (self_preservation, #16 of 17).

Compare this model

Only models sharing at least one published sub-eval are listed.

Official and reference links

Published eval results

Rank is within that sub-eval. Black marks the observed result; the grey dot marks the value implied by the global rank. Values stay on each source’s native scale.

Eval / sub-evalRankValueDistributionBetterSource
CRiskEvaldeception_willingness#14 / 1735.09↓ lowerSource ↗official
CRiskEvaldesire_for_resource#16 / 1749.23↓ lowerSource ↗official
CRiskEvalharmful_goal#16 / 1755.31↓ lowerSource ↗official
CRiskEvalimprovement_intent#16 / 1757.29↓ lowerSource ↗official
CRiskEvalmalicious_coordination#15 / 1729.04↓ lowerSource ↗official
CRiskEvalself_preservation#16 / 1744.89↓ lowerSource ↗official
CRiskEvalsituational_awareness#16 / 1769.14↓ lowerSource ↗official