List your company
ResearchFormerly WizWand

Logical Reasoning on BBH Web of Lies

98Accuracyevaluation-instructed prompt optimization
64.7273.368290.64Nov 25, 2025
Updated 5mo ago

Evaluation results

MethodLinks
evaluation-instructed prompt optimizationBackbone=GPT-4o, Optimization Type=Query-dependent2025.1198
LLM onlyBackbone=GPT-4o, Optimization Type=None2025.1196
Pro-RefineBackbone=GPT-4o, Optimization Type=Query-dependent2025.1196
TextGradBackbone=GPT-4o, Optimization Type=Template2025.1196
APEBackbone=GPT-4o, Optimization Type=Template2025.1195
Self-RefineBackbone=GPT-4o, Optimization Type=Query-dependent2025.1194
TextGradBackbone=LLaMA-3, Optimization Type=Template2025.1174
TextGradBackbone=LLaMA-3.1, Optimization Type=Template2025.1173
APEBackbone=LLaMA-3, Optimization Type=Template2025.1172
Pro-RefineBackbone=LLaMA-3, Optimization Type=Query-dependent2025.1171
APEBackbone=LLaMA-3.1, Optimization Type=Template2025.1171
Pro-RefineBackbone=LLaMA-3.1, Optimization Type=Query-dependent2025.1170
evaluation-instructed prompt optimizationBackbone=LLaMA-3, Optimization Type=Query-dependent2025.1169
LLM onlyBackbone=LLaMA-3.1, Optimization Type=None2025.1169
evaluation-instructed prompt optimizationBackbone=LLaMA-3.1, Optimization Type=Query-dependent2025.1168
Self-RefineBackbone=LLaMA-3, Optimization Type=Query-dependent2025.1167
LLM onlyBackbone=LLaMA-3, Optimization Type=None2025.1166
Self-RefineBackbone=LLaMA-3.1, Optimization Type=Query-dependent2025.1166