LLMs4Science

Introduction

Rapid advances in large language models (LLMs) provide an unprecedented opportunity to further scientific inquiry across scientific disciplines and domains. Despite remarkable feats in natural language tasks often exclusively indicative of human intelligence, the potential of LLMs beyond natural language has yet to be realized. Outstanding challenges in scientific disciplines, such as molecular biology, material science, climate science, geology, hydrology, and various domains within them, such as drug discovery, quantum material design, weather forecasting, and more, necessitate the integration of heterogeneous, multi-modal datasets resulting from diverse physical processes, as well as the injection of deep domain knowledge accumulated over decades of discovery about the inherent physical processes that govern the natural and biological world.

Objective: Through this workshop we want to bring together diverse communities of researchers for a thoughtful and concerted effort at advancing research on LLMs and their integration in learning-enabled frameworks. The hope is that by bringing together diverse researchers, we will properly formulate problem spaces and the proper datasets, standards, and benchmarks of scientific progress to support and more, excitingly, to spur scientific breakthroughs across diverse scientific disciplines and domains. The proposed workshop will be an important medium to leverage synergies and facilitate an open exchange of ideas so that the research community can advance both foundational research in LLMs, inspired beyond the classNameic NLP tasks and perspectives, as well as science-inspired research on LLM-enabled scientific breakthroughs across a variety of scientific disciplines and domains.