List your company

Causal Tool Use on Causal-ToolBench Adversarial confounding

0False Execution RateOracleSCM
-1.8210.46522.7535.035May 9, 2026

Evaluation results

MethodLinks
OracleSCMNumber of seeds=72026.05054.51002.76
CIVeXNumber of seeds=72026.05039.484.92.23
CausalPromptingCoT-OpusBackbone=Opus, Number of seeds=52026.0502065.21.65
AlwaysAbstainNumber of seeds=72026.050045.50.99
CausalNoExperimentNumber of seeds=72026.050045.50.99
ContextOnlyNoCausalNumber of seeds=72026.050045.50.99
ObservationalAssociationNumber of seeds=72026.050045.50.99
CausalPromptingCoT-SonnetBackbone=Sonnet, Number of seeds=52026.0512872.21.82
NameOnlyClassifierNumber of seeds=72026.058.87.944.50.78
CausalPromptingTerseNumber of seeds=72026.0510.239.474.61.57
PolicyGateNumber of seeds=72026.0528.36.924-0.27
FamilyMajorityClassifierNumber of seeds=72026.0545.554.554.5-0.03
SchemaGateNumber of seeds=72026.0545.554.554.5-0.03
SemanticOntologyGateNumber of seeds=72026.0545.554.554.5-0.03