AI and the World's Hard Problems · Chemistry & materials · arXiv · 2026-10-26
PULSES: multi-agent AI for autonomous excited-state model building and dynamics
Simulating molecular excited states requires tracking electronic and nuclear dynamics, but direct quantum-chemistry calculations are costly and limit system size and timescales. PULSES is a multi-agent system that automates building and deploying machine-learning models for this work within boundaries approved by an expert, and it was evaluated on 15 computational chemistry tasks. It was applied to perylene exciton Hamiltonians, solvent-dependent biphenyl relaxation, and excited-state proton transfer in 10-hydroxybenzo[h]quinoline, producing traceable, expert-bounded excited-state models while keeping scientific validation as a requirement.