Reproduce “Integrating Large Language Models with Graph-based Reasoning for Conversational Question Answering” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Parag Jain, Mirella Lapata
Researchers at the University of Edinburgh developed a framework that integrates Large Language Models with graph-based reasoning to enhance conversational question answering across heterogeneous information sources. The approach uses a graph-based representation of diverse evidence and an evidence memory system, demonstrating improved performance on the ConvMix dataset compared to LLMs without such integration, especially in multi-turn dialogues.
Key insights
- Graph-based representations effectively integrate heterogeneous information sources (knowledge graphs, text, tables, infoboxes) for Large Language Models.
- An evidence memory module allows the system to track and reuse previously retrieved information, enhancing performance in multi-turn conversations.
- Parameter-efficient fine-tuning (LoRA) can effectively integrate structured graph knowledge with LLM capabilities for complex reasoning tasks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.