Shang Hong Sim

I’m a graduating Master’s student at the DeCLaRE Lab, advised by Prof. Soujanya Poria. My research focuses on addressing core limitations of Large Language Models (LLMs), particularly hallucinations, as they increasingly serve as information consolidators 🧠. I developed a novel metric to evaluate LLM groundedness in Retrieval-Augmented Generation (RAG) systems and proposed a preference dataset to align models for improved grounding 🎯.

My prior work spans interdisciplinary applications, including leveraging Unity3D to bridge vision and language tasks 👀 and utilizing bioinformatics to identify therapeutically significant gene sets 🧬. With experience in state-of-the-art NLP and cancer biology, I hope to contribute experimental and analytical expertise to the projects that I am part of. I am passionate about advancing cutting-edge research and translating complex findings into practical, impactful solutions that drive both technical and business innovation 💡.

selected publications

Measuring and Enhancing Trustworthiness of LLMs in RAG through Grounded Attributions and Learning to Refuse

Maojia Song^*, Shang Hong Sim^*, Rishabh Bhardwaj, and 3 more authors

2024
Evaluating the Generation of Spatial Relations in Text and Image Generative Models

Shang Hong Sim^*, Clarence Lee^*, Alvin Tan, and 1 more author

2024
p53-dependent crosstalk between DNA replication integrity and redox metabolism mediated through a NRF2-PARP1 axis

Gamal Ahmed Elfar, Obed Aning, Tsz Wai Ngai, and 13 more authors

Nucleic Acids Research, Sep 2024

Abs DOI

Mechanisms underlying p53-mediated protection of the replicating genome remain elusive, despite the quintessential role of p53 in maintaining genomic stability. Here, we uncover an unexpected function of p53 in curbing replication stress by limiting PARP1 activity and preventing the unscheduled degradation of deprotected stalled forks. We searched for p53-dependent factors and elucidated RRM2B as a prime factor. Deficiency in p53/RRM2B results in the activation of an NRF2 antioxidant transcriptional program, with a concomitant elevation in basal PARylation in cells. Dissecting the consequences of p53/RRM2B loss revealed a crosstalk between redox metabolism and genome integrity that is negotiated through a hitherto undescribed NRF2-PARP1 axis, and pinpoint G6PD as a primary oxidative stress-induced NRF2 target and activator of basal PARylation. This study elucidates how loss of p53 could be destabilizing for the replicating genome and, importantly, describes an unanticipated crosstalk between redox metabolism, PARP1 and p53 tumor suppressor pathway that is broadly relevant in cancers and can be leveraged therapeutically.