List your company

Multi-hop Question Answering on MuSiQue (Exact Match, RAG Improvement)

16.9Exact MatchPyRAG
1.6125.5819.5513.519May 13, 2026

Evaluation results

MethodLinks
PyRAGBackbone=Qwen2.5-72B-Instruct2026.0516.96.3
IRCoTBackbone=Qwen2.5-72B-Instruct2026.0514.2—
ITER-RETGENBackbone=Qwen2.5-72B-Instruct2026.0513.8—
PyRAGBackbone=Qwen2.5-7B-Instruct2026.0511.87.1
Self-AskBackbone=Qwen2.5-72B-Instruct2026.0511.8—
Vanilla RAGBackbone=Qwen2.5-72B-Instruct2026.0510.6—
CoTBackbone=Qwen2.5-72B-Instruct2026.059.9—
IRCoTBackbone=Qwen2.5-7B-Instruct2026.059.1—
ITER-RETGENBackbone=Qwen2.5-7B-Instruct2026.058.7—
Self-AskBackbone=Qwen2.5-7B-Instruct2026.056.8—
Direct InferenceBackbone=Qwen2.5-72B-Instruct2026.055.5—
Vanilla RAGBackbone=Qwen2.5-7B-Instruct2026.054.7—
Direct InferenceBackbone=Qwen2.5-7B-Instruct2026.053.1—
CoTBackbone=Qwen2.5-7B-Instruct2026.052.2—