List your company
ResearchFormerly WizWand

Multi-hop reasoning on MuSiQue Closed Book (test)

21.14Accuracy/F1DCO (Ours)
3.32487.949912.57517.2001Jun 2, 2026
Updated 5mo ago

Evaluation results

MethodLinks
DCO (Ours)Model=Llama-3-70B, Chain-of-Thought=With CoT2026.0621.14
DeCoRe (static)Model=Llama-3-70B, Chain-of-Thought=With CoT2026.0620.6
ITIModel=Llama-3-70B, Chain-of-Thought=With CoT2026.0620.44
ADModel=Llama-3-70B, Chain-of-Thought=With CoT2026.0620.23
GreedyModel=Llama-3-70B, Chain-of-Thought=With CoT2026.0620.15
DoLaModel=Llama-3-70B, Chain-of-Thought=With CoT2026.0620.15
CDModel=Llama-3-70B, Chain-of-Thought=With CoT2026.0617.17
DoLaModel=Llama-3-8B, Chain-of-Thought=With CoT2026.0614.94
DeCoRe (static)Model=Llama-3-8B, Chain-of-Thought=With CoT2026.0614.69
GreedyModel=Llama-3-8B, Chain-of-Thought=With CoT2026.0614.61
DCO (Ours)Model=Llama-3-8B, Chain-of-Thought=With CoT2026.0614.56
ADModel=Llama-3-8B, Chain-of-Thought=With CoT2026.0614.4
DCO (Ours)Model=Llama-3-70B, Chain-of-Thought=Without CoT2026.0611.96
GreedyModel=Llama-3-70B, Chain-of-Thought=Without CoT2026.0611.79
DeCoRe (static)Model=Llama-3-70B, Chain-of-Thought=Without CoT2026.0611.79
DoLaModel=Llama-3-70B, Chain-of-Thought=Without CoT2026.0611.42
ADModel=Llama-3-70B, Chain-of-Thought=Without CoT2026.0611.38
CDModel=Llama-3-70B, Chain-of-Thought=Without CoT2026.0610.92
ITIModel=Llama-3-70B, Chain-of-Thought=Without CoT2026.0610.88
DeCoRe (static)Model=Llama-3-8B, Chain-of-Thought=Without CoT2026.067.9
GreedyModel=Llama-3-8B, Chain-of-Thought=Without CoT2026.067.41
DCO (Ours)Model=Llama-3-8B, Chain-of-Thought=Without CoT2026.067.41
DoLaModel=Llama-3-8B, Chain-of-Thought=Without CoT2026.067.24
ADModel=Llama-3-8B, Chain-of-Thought=Without CoT2026.066.99
ITIModel=Llama-3-8B, Chain-of-Thought=With CoT2026.064.18
ITIModel=Llama-3-8B, Chain-of-Thought=Without CoT2026.064.01