AI-enhanced Research explores how large language models and autonomous agents can assist scientific discovery by planning simulations, coordinating domain tools, and improving research code. The focus is on building reliable agentic workflows that remain grounded in physics, chemistry, and numerical constraints rather than treating scientific computing as a generic text-generation problem.
A representative direction in this effort is FermiLink, a unified agent framework for multidomain autonomous scientific simulations (Meng et al., 2026). This line of work studies how LLM-based agents can decompose scientific tasks, interface with simulation backends, iteratively refine code, and organize multi-step computational experiments across different research settings.
