List your company
ResearchFormerly WizWand

Multi-Hop Question Answering on Synthetic Multi-Hop Benchmark 2-hop

1Average F1 ScoreChain-of-Thought (CoT)
0.14720.36860.590.8114Sep 25, 2025
Updated 8mo ago

Evaluation results

MethodLinks
Chain-of-Thought (CoT)Context Length=0.5k, Backbone=Qwen3-14B2025.091
InfoQAContext Length=0.5k, Backbone=Qwen3-14B2025.091
InfoQAContext Length=Average, Backbone=Qwen3-14B2025.090.97
ReActContext Length=2k, Backbone=Qwen3-8B2025.090.93
InfoQAContext Length=10k, Backbone=Qwen3-14B2025.090.89
Plan-and-Solve (P&S)Context Length=4k, Backbone=Qwen3-8B2025.090.78
DirectContext Length=0.5k, Backbone=Qwen3-14B2025.090.78
InfoQAContext Length=10k, Backbone=Qwen3-8B2025.090.72
Self-Ask (S-A)Context Length=8k, Backbone=Qwen3-8B2025.090.67
DirectContext Length=Average, Backbone=Qwen3-14B2025.090.52
DirectContext Length=10k, Backbone=Qwen3-14B2025.090.18