List your company

Multi-hop Question Answering on 2WikiMultihopQA (Exact Match/RAG Improvement)

44.4Exact Match (EM)PyRAG
9.45618.52827.636.672May 13, 2026

Evaluation results

MethodLinks
PyRAGBackbone=Qwen2.5-72B-Instruct2026.0544.414
ITER-RETGENBackbone=Qwen2.5-72B-Instruct2026.0540.2—
IRCoTBackbone=Qwen2.5-72B-Instruct2026.0535.8—
PyRAGBackbone=Qwen2.5-7B-Instruct2026.0533.414.5
Self-AskBackbone=Qwen2.5-72B-Instruct2026.0532.5—
Vanilla RAGBackbone=Qwen2.5-72B-Instruct2026.0530.4—
ITER-RETGENBackbone=Qwen2.5-7B-Instruct2026.0528.9—
CoTBackbone=Qwen2.5-72B-Instruct2026.0525.4—
IRCoTBackbone=Qwen2.5-7B-Instruct2026.0524.8—
Self-AskBackbone=Qwen2.5-7B-Instruct2026.0521.5—
Direct InferenceBackbone=Qwen2.5-72B-Instruct2026.0520.6—
Vanilla RAGBackbone=Qwen2.5-7B-Instruct2026.0518.9—
Direct InferenceBackbone=Qwen2.5-7B-Instruct2026.0512.6—
CoTBackbone=Qwen2.5-7B-Instruct2026.0510.8—