List your company

Instruction Hierarchy Adherence on IHEval Aligned Single-turn

84.8AccuracyVanilla
40.371251.905663.4474.9744Jun 29, 2026

Evaluation results

MethodLinks
VanillaModel=Qwen3-8B-Instruct2026.0684.8
VerIHModel=Qwen3-8B-Instruct2026.0684.59
PromptingModel=Qwen3-8B-Instruct2026.0683.6
IHDec + VerIHModel=Qwen3-8B-Instruct2026.0681.8
IHDec + PromptingModel=Qwen3-8B-Instruct2026.0678.35
IHDecModel=Qwen3-8B-Instruct2026.0677.2
IHDecModel=Qwen3-4B-Instruct2026.0676.7
PromptingModel=Qwen3-4B-Instruct2026.0675.9
IHDec + PromptingModel=Qwen3-4B-Instruct2026.0675.9
IHDecModel=Llama-3.1-8B-Instruct2026.0673.53
VanillaModel=Qwen3-4B-Instruct2026.0671
VanillaModel=Llama-3.1-8B-Instruct2026.0666.05
IHDec + PromptingModel=Llama-3.1-8B-Instruct2026.0661.8
PromptingModel=Llama-3.1-8B-Instruct2026.0656.5
IHDec + ISEModel=Llama-3.1-8B-Instruct2026.0655.1
ISEModel=Llama-3.1-8B-Instruct2026.0642.08