Abstract. We provide a mirror symmetry theorem in a range of cases where the state-of-the-art techniques relying on concavity or convexity do not apply. More specifically, we work on a family of FJRW potentials named after Fan, Jarvis, Ruan, and Witten's quantum singularity theory and viewed as the counterpart of a non-convex Gromov-Witten potential via the physical LG/CY correspondence. The main result provides an explicit formula for Polishchuk and Vaintrob's virtual cycle in genus zero. In the non-concave case of the so-called chain invertible polynomials, it yields a compatibility theorem with the FJRW virtual cycle and a proof of mirror symmetry for FJRW theory. In the last two decades, mirror symmetry has been a central statement in theoretical physics and a fundamental driving force for several developments in mathematics. For instance it can be phrased mathematically as a prediction on GromovWitten invariants, namely the intersection numbers attached to curves traced on a Calabi-Yau variety. In this form, it has been proven in a vast range of concrete cases: the most famous example provides a full computation of the genus-zero invariants enumerating rational curves on the quintic threefold [21,29]. Even in this case, it is often pointed out how we still lack a complete computation in higher genus (only the genus-one case was completely proven by Zinger [35]).But even in genus zero the problem of computing Gromov-Witten invariants of projective varieties is far from being completely solved; indeed, most known techniques focus on computing Gromov-Witten invariants attached to cohomology classes which lie in the so-called ambient part of cohomology: the restriction to classes from the ambient projective space. For the quintic threefold, working with ambient cohomology classes turns out to determine the entire theory; however, in general, this scheme covers only a tiny portion of quantum cohomology. Remarkably, even the ambient cohomology classes may pose problem as soon as we work with orbifolds.It is interesting to notice that these gaps in Gromov-Witten computation all arise from the same phenomena: as we argue below, certain positivity or negativity conditions named convexity and concavity are not always satisfied, making the virtual cycle 1 challenging to compute. In genus one, this difficulty was overcome by Zinger after a great deal of hard work, but we still lack a comprehensive approach for higher genus. Guided by the frame of ideas of mirror symmetry and the Landau-Ginzburg/Calabi-Yau correspondence, we switch to the quantum theory of singularities introduced by Fan, Jarvis, and Ruan [17,18] based on ideas of Witten [34] (FJRW theory). Single non-concave quantum invariants were inferred from concave ones using tautological relation (e.g. using WDVV equation as in [16,17,26]), but so far no systematic approach tackling directly the virtual cycle has been taken. Polishchuk and Vaintrob recently opened the way to an algebraic computation: their construction [32] of a virtual cycle is given by applyin...