MPR-CiteG Framework Achieves Second Place in ScienceON AI Challenge
A research paper introduces MPR-CiteG, a framework that secured second place in the ScienceON AI Challenge. It tackles inefficient retrieval and lack of source verification in generative AI. The system comprises a Multi-Portfolio Retriever (MPR) for diverse information retrieval and a Citation-Grounded Generation (CiteG) module to ensure factual consistency and source attribution. Experiments on the challenge dataset validate its effectiveness. The code is available on GitHub.
Key facts
- MPR-CiteG achieved second place in the ScienceON AI Challenge.
- The framework addresses inefficient retrieval and absence of source verification.
- It consists of Multi-Portfolio Retriever (MPR) and Citation-Grounded Generation (CiteG).
- MPR retrieves diverse and relevant information.
- CiteG ensures factual consistency and explicit source attribution.
- Experiments on the challenge dataset validate the approach.
- Code is available at https://github.com/.
- The paper is on arXiv with ID 2607.22706.
Entities
Institutions
- ScienceON
- arXiv