Authors: Le Zhang, Yihong Wu

Published on: January 12, 2024

Impact Score: 8.07

Arxiv code: Arxiv:2401.06311

Summary

• What is new: Introduction of MuGI, a new framework to augment IR by generating multiple pseudo-references.
• Why this is important: Limitations of existing query expansion strategies in information retrieval.
• What the research proposes: MuGI framework uses LLMs to generate pseudo-references for enhancing query retrieval without additional training.
• Results: Improvement of over 18% in BM25 on TREC DL and 7.5% on BEIR, surpassing larger models with a smaller 110M parameter retriever.

Technical Details

Technological frameworks used: MuGI (Multi-Text Generation Integration)

Models used: LLMs, ANCE, DPR, BM25

Data used: TREC DL dataset, BEIR benchmark

Potential Impact

Search engine providers, content management systems, digital libraries, enterprise data retrieval companies

Want to implement this idea in a business?

We have generated a startup concept here: SmartSearchAI.