Information content is a general-purpose collective variable for enhanced sampling

Enhanced sampling methods typically require predefined collective variables (CVs) that presuppose knowledge of reaction coordinates, restricting the discovery of unanticipated transition mechanisms or intermediates. Here, we show that a local measure of information content in atomistic systems is a general-purpose CV for rare event sampling across molecular and condensed-phase systems. The method biases simulations toward information-changing configurations following a well-tempered metadynamics approach, thus balancing novelty and thermodynamic accessibility. Blind exploration of potential energy surfaces enables unsupervised discovery of metastable basins and reaction pathways, including competing transition channels inaccessible to conventional order parameters. We demonstrate the generality of the method across five systems spanning conformational sampling, homogeneous nucleation, glass formation, and solid-state phase transformations. Our approach uncovers non-classical phase transformation mechanisms despite its focus on general information theoretical quantities rather than kinetic or thermodynamic factors.