Professor Scott Reed, from CU Denver's Chemistry Department, introduced a groundbreaking approach to improving artificial intelligence (AI) reliability in chemical research. Large Language Models (LLMs) have the potential to assist scientists, but they often generate misleading or incorrect information, a challenge known as "hallucination."
In his latest study, Professor Reed demonstrates how LLMs can produce more accurate chemical predictions without the need for costly retraining. By combining Retrieval-Augmented Generation (RAG) with machine learning optimized prompting techniques, his research improves AI performance on key molecular properties used in drug discovery.
This work represents a major step toward making AI a more dependable tool for chemistry and scientific research at-large.