Lattice networks with dissipative interactions can be used to describe the mechanics of discrete mesostructures of materials such as 3D-printed structures and foams. This contribution deals with the crack initiation and propagation in such materials and focuses on an adaptive multiscale approach that captures the spatially evolving fracture. Lattice networks naturally incorporate non-locality, large deformations and dissipative mechanisms taking place inside fracture zones. Because the physically relevant length scales are significantly larger than those of individual interactions, discrete models are computationally expensive. The Quasicontinuum (QC) method is a multiscale approach specifically constructed for discrete models. This method reduces the computational cost by fully resolving the underlying lattice only in regions of interest, while coarsening elsewhere. In this contribution, the (variational) QC is applied to damageable lattices for engineering-scale predictions. To deal with the spatially evolving fracture zone, an adaptive scheme is proposed. Implications induced by the adaptive procedure are discussed from the energy-consistency point of view, and theoretical considerations are demonstrated on two examples. The first one serves as a proof of concept, illustrates the consistency of the adaptive schemes and presents errors in energies. The second one demonstrates the performance of the adaptive QC scheme for a more complex problem. AN ADAPTIVE VARIATIONAL QUASICONTINUUM METHODOLOGY FOR LATTICE 175 beam (fibre or yarn) versus that of the network. Second, the formulation and implementation of lattice models is generally significantly easier compared with that of alternative continuum models. Large deformations, large yarn reorientations and fracture are easier to formulate and implement (cf. e.g. the continuum model of Peng and Cao [5] that deals with large yarn reorientations). Thanks to the simplicity and versatility of lattice networks, they are furthermore used for the description of heterogeneous cohesive-frictional materials such as concrete. The reason is that discrete models can realistically represent distributed microcracking with gradual softening, implement material structure with inhomogeneities, capture non-locality of damage processes and reflect deterministic or stochastic size effects. Examples of the successful use of lattice models for such materials are given in [6][7][8][9].As lattice models are typically constructed at the meso-scale, micro-scale or nano-scale, they require reduced-model techniques to allow for application-scale simulations. A prominent example is the Quasicontinuum (QC) method, which specifically aims at discrete lattice models. The QC method was originally introduced for conservative atomistic systems by Tadmor et al. [10] and extended in numerous aspects later on, see for example [11][12][13]. Subsequent generalizations for lattices with dissipative interactions (e.g. plasticity and bond sliding) were provided in [14,15]. In principle, the QC is a numerical p...