Research map / Large language models / Retrieval and knowledge
Distill-SynthKG: Distilling Knowledge Graph Synthesis Workflow for Improved Coverage and Efficiency
Prafulla Kumar Choubey, Xin Su, Man Luo, XIANGYU PENG, Caiming Xiong, Tiep Le, Shachar Rosenman, Vasudev Lal, Phil L Mui, Ricky Ho, Phillip Howard, Chien-Sheng Wu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use multi-step SynthKG pipeline instead of single-step prompting with Llama-3-8b | Llama-3-8b KG construction · MuSiQue KG coverage | triplet coverage: 32.09 -> 38.84 | Improved |
| use multi-step SynthKG pipeline instead of single-step prompting with Llama-3-70b | Llama-3-70b KG construction · 2Wiki KG coverage | triplet coverage: 54.10 -> 63.30 | Improved |
| distill SynthKG into single-step 8b model (D-SynthKG-8b) versus untrained Llama-3-8b | Llama-3-8b · MuSiQue KG coverage | triplet coverage: 32.09 -> 46.90 | Improved |
| distilled 8b model versus its multi-step 70b teacher pipeline | D-SynthKG-8b · 2Wiki KG coverage | triplet coverage: 63.30 -> 58.27 | Hurt |
| use Graph+LLM retriever on D-SynthKG-8b KG versus the Llama-3-8b KG | Graph+LLM retrieval · MuSiQue retrieval | Hits@2: 31.33 -> 53.35 | Improved |
| use Graph+LLM on D-SynthKG-8b KG versus Llama-3-8b KG for multi-hop QA | Graph+LLM QA · Multi-hop QA average | EM: 0.281 -> 0.433 | Improved |
| add LLM re-ranking on top of dense retrieval (Dense+LLM) | Dense retriever · Multi-hop QA average | EM: 0.363 -> 0.394 | Improved |
| use proposition-entity Graph+LLM retriever instead of dense retrieval | retrieval · MuSiQue, 2Wiki, HotpotQA retrieval | Hits@2: +12.75 points over dense, +1.67 over Dense+LLM | Improved |
| retrieve propositions instead of triples (ablation) | KG retrieval source · multi-hop QA retrieval | Hits@10: +0.89 Hits@10 | Improved |
| use full graph-based retrieval instead of propositions only (ablation) | KG retrieval source · multi-hop QA retrieval | Hits@10: +2.50 | Improved |
| add propositions to retrieved chunks as QA context (ablation) | RAG context · multi-hop QA | EM: +2 EM | Improved |
| add 2-hop paths to retrieved chunks as QA context (ablation) | RAG context · multi-hop QA | EM: +2.7 EM | Improved |
Closest papers by meaning
- A Systematic Exploration of Knowledge Graph Alignment with Large Language Models in Retrieval Augmented Generation (AAAI 2025 · similarity 0.91)
- Think-on-Graph 2.0: Deep and Faithful Large Language Model Reasoning with Knowledge-guided Retrieval Augmented Generation (ICLR 2025 · similarity 0.90)
- Diagnosing and Addressing Pitfalls in KG-RAG Datasets: Toward More Reliable Benchmarking (NeurIPS 2025 · similarity 0.89)
- Simple is Effective: The Roles of Graphs and Large Language Models in Knowledge-Graph-Based Retrieval-Augmented Generation (ICLR 2025 · similarity 0.89)
- KGGen: Extracting Knowledge Graphs from Plain Text with Language Models (NeurIPS 2025 · similarity 0.89)
- Knowledge Graph Prompting for Multi-Document Question Answering (AAAI 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
