List your company

Multi-hop Question Answering on HotpotQA (Expected utility)

88.4Expected UtilityFlowCompile
56.88865.06973.2581.431May 13, 2026

Evaluation results

MethodLinks
FlowCompileWorkflow Category=Main Method2026.0588.4
Pref-Aware MaASWorkflow Category=Workflow with Router2026.0578.7
MaAS (Qwen3-4B)Workflow Category=Workflow with Router2026.0574.2
KNN RouterWorkflow Category=Workflow with Router2026.0574.1
Pref-Aware RouterWorkflow Category=Workflow with Router2026.0569.3
Qwen3-4BWorkflow Category=Fixed Workflow2026.0568.3
Qwen3-32BWorkflow Category=Single Model2026.0567.4
MaAS (Qwen3-8B)Workflow Category=Workflow with Router2026.0565.2
QwQ-32BWorkflow Category=Single Model2026.0562.6
Qwen3-8BWorkflow Category=Fixed Workflow2026.0558.1