We study attractively interacting spin-1 2 fermions on the square lattice subject to a spin population imbalance. Using unbiased diagrammatic Monte Carlo simulations we find an extended region in the parameter space where the Fermi liquid is unstable towards formation of Cooper pairs with nonzero center-of-mass momentum, known as the Fulde-Ferrell-Larkin-Ovchinnikov (FFLO) state. In contrast to earlier mean-field and quasi-classical studies we provide quantitative and well-controlled predictions on the existence and location of the relevant Fermi-liquid instabilities. The highest temperature where the FFLO instability can be observed is about half of the superfluid transition temperature in the unpolarized system. PACS numbers: 71.10. Hf,37.10.Jk,74.20.Mn Fifty years after the initial prediction by Fulde, Ferrell, Larkin, and Ovchinnikov (FFLO) [1,2], superconducting phases with spontaneously broken translational invariance are still at the center of interest in such diverse fields as solid state physics, cold atomic gases, nuclear physics, and dense quark matter in neutron stars [3][4][5][6][7]. While the underlying mechanism is generic enough to apply to any partially polarized Fermi system, it has proven surprisingly difficult to unambiguously observe such phases in nature. Recently, however, experimental evidence has been mounting for their existence in heavy fermion compounds [8][9][10] and layered organic materials [11][12][13][14][15]. On the other hand, experiments with ultracold atoms, which are among the cleanest imbalanced Fermi systems without the need for a magnetic field, so far failed to demonstrate inhomogeneous superfluidity [16,17] -although there is some evidence for such a phase in one dimension (1D) [18] -possibly due to small extent of the parameter region where an FFLO phase may exist in three dimensions (3D) and difficulties in reaching sufficiently low temperatures [5].On the theoretical side, results on the existence and nature of FFLO phases based on well-controlled microscopic theories are scarce, with the exception of 1D systems, where exact analytical and numerical studies are possible [19][20][21][22][23][24][25], and where finite-momentum pairing is a generic feature of the spin-imbalanced phase diagram. In higher dimensions, most studies are based on effective field theories in the neighborhood of critical points or resort to quasi-classical or mean-field approximations. For 3D Fermi gases, many features of the mean-field phase diagram [26] have been corroborated by fixed-node diffusion quantum Monte Carlo calculations [27]; whether the FFLO phase does exist in a small sliver of the phase diagram, as predicted by the mean-field theory, is however still subject to debate.The FFLO state is expected [3,4] to occupy a larger parameter region in two dimensions (2D), and lattice effects may further increase its stability [28,29]. Correspondingly, mean-field calculations [30,31] and realspace dynamical mean-field theory (DMFT) for fermions in anisotropic optical lattices find a st...