The Research and Postgraduate Centre congratulates Feng Junzhe, an alumnus from Electrical and Electronics Engineering programme, Asst. Prof. Ts. Dr. Kang Chia Chao and team from the School of Artificial Intelligence and Robotics and S.M.A.R.T. Nexus Centre of Excellence, on the publication of their recent research article in Scientific Reports, an SCIE Q1 journal with a latest impact factor of 4.9.

Feng Junzhe is the first author of the article. Titled “Fine grained reranking via caption bridging for knowledge augmented visual question answering”, the research focuses on solving the granularity mismatch problem within Retrieval-Augmented Generation (RAG) frameworks for knowledge-intensive reasoning. Driven by the limitations of traditional RAG systems where coarse-grained retrieval units (like entire passages or global images) fail to align with the highly specific reasoning requirements of generation tasks, there is a critical need for localized visual evidence. To overcome this, the researchers engineered a Fine-Grained Retrieval-Augmented Generation (FG-RAG) framework that enhances the standard Contrastive Language–Image Pre-training (CLIP) architecture with patch-level contrastive supervision, allowing for explicit alignment between localized image regions and their corresponding textual fragments. 

Furthermore, the team introduced a joint retrieval-reranking optimization mechanism that unifies a dense retriever with a large language model (LLM)-based reranker using a shared relevance loss and a novel Score Alignment Strategy. By creating bidirectional feedback between retrieval precision and generation quality, this research maps out how to significantly enhance factual grounding and contextual coherence, reducing hallucination rates by 15 percentage points compared to conventional RAG systems.

The published full article can be accessed at DOI: 10.1038/s41598-026-57349-w

This work is financially supported by Xiamen University Malaysia Research Fund (XMUMRF/2025-C15/IECE/0059).

Asst. Prof. Ts. Dr. Kang Chia Chao research lies at the cutting edge of sustainability, focusing on renewable energy, 5G/6G wireless communication, healthcare for sustainable development. Dr. Kang is currently seeking highly motivated Master's students to join his dynamic research team. If you are eager to work on impactful, sustainable technologies, please contact him directly at chiachao.kang@xmu.edu.my.