List your company

Instruction Hierarchy Adherence on IHEval Conflict Single-turn

52.13AccuracyIHDec
13.6523.6433.6343.62Jun 29, 2026

Evaluation results

MethodLinks
IHDecModel=Llama-3.1-8B-Instruct2026.0652.13
IHDec + PromptingModel=Qwen3-4B-Instruct2026.0651.1
IHDec + ISEModel=Llama-3.1-8B-Instruct2026.0650.4
IHDec + VerIHModel=Qwen3-8B-Instruct2026.0649.3
IHDec + PromptingModel=Qwen3-8B-Instruct2026.0649.1
PromptingModel=Qwen3-8B-Instruct2026.0648.9
VerIHModel=Qwen3-8B-Instruct2026.0647.56
IHDecModel=Qwen3-4B-Instruct2026.0645.7
IHDecModel=Qwen3-8B-Instruct2026.0643.8
PromptingModel=Qwen3-4B-Instruct2026.0639.7
IHDec + PromptingModel=Llama-3.1-8B-Instruct2026.0637.9
VanillaModel=Qwen3-4B-Instruct2026.0631.3
VanillaModel=Qwen3-8B-Instruct2026.0627.7
ISEModel=Llama-3.1-8B-Instruct2026.0626.83
PromptingModel=Llama-3.1-8B-Instruct2026.0615.7
VanillaModel=Llama-3.1-8B-Instruct2026.0615.13