Reinforcement Coding
BlogPlaygroundTagsAbout

Playground

Interactive demos and experiments. Explore AI, ML, and web technologies in action.

Multi-Hop RAG

Compare adaptive multi-hop retrieval techniques on MuSiQue dataset. See how iterative reasoning with query decomposition and reranking outperforms simple RAG.

Features

  • Simple RAG baseline
  • Adaptive multi-hop (1-3 hops)
  • Query expansion + rank fusion
  • 150 sample questions (2-3 hop)
MuSiQue
Dataset
7.6K
Paragraphs
71K
Records
+30%
Accuracy Gain
RAGLLMVector SearchDSPy
Launch Demo
mailMailgithubGitHublinkedinLinkedin
Xiaoyi Zhu
•
© 2025
•
Reinforcement Coding