List your company

Multi-hop Question Answering on MuSiQue (EM, Retrieval Rate, Latency)

15.2Exact Match (EM)TARG-Margin
8.75210.42612.113.774Nov 12, 2025

Evaluation results

MethodLinks
TARG-MarginBackbone=Qwen2.5-7B2025.1115.21.60.16
Never-RAGBackbone=Qwen2.5-7B2025.1114.600.18
Self-Eval GateBackbone=Qwen2.5-7B2025.1114.624.65.66
CRAG-liteBackbone=Qwen2.5-7B2025.1114.2231.76
SKRBackbone=Qwen2.5-7B2025.1114.213.20.92
Self-RAG-liteBackbone=Qwen2.5-7B2025.1113.448.21.8
Always-RAGBackbone=Qwen2.5-7B2025.111010012.24
FLAREBackbone=Qwen2.5-7B, θ=0.52025.119.81004.73
FLAREBackbone=Qwen2.5-7B, θ=0.32025.11998.817.86