First-principles calculations are carried out to study the effect of electron correlations on relative structural stability, magnetism, and spin-dependent transport in CeMnNi 4 intermetallic compound. The correct description of Coulomb repulsion of Mn 3d electrons is shown to play a crucial role in reproducing the experimentally observed cubic phase of CeMnNi 4 as well as its relatively high degree of transport spin polarization ͑ϳ66%͒. These are the two fundamental properties of this compound which conventional density-functional theory approaches fail to predict correctly. The reason for this failure is attributed to an extreme overdelocalization of Mn 3d charges causing a strong d-d hybridization between Mn and Ni atoms in the orthorhombic phase. Such an artificial hybridization, in turn, lowers the relative total energy of the orthorhombic phase with respect to the cubic one. It also leads to an incorrect carrier concentration and mobility at the Fermi level and, consequently, yields much lower degree of transport spin polarization for this nearly half-metallic compound.Electron correlation plays a crucial role in understanding the electronic structure and magnetism in narrow-band systems. Density-functional theory ͑DFT͒ within its localdensity approximation ͑LDA͒ and generalized gradient approximation ͑GGA͒ turns out to be impressively successful in describing the ground-state properties of many materials. However, as an approximation, LDA/GGA faces serious difficulties for the strongly correlated systems, e.g. hightemperature oxide superconductors 1 and the middle-to-late transition-metal oxides. 2 Such a complication arises from the fact that the current implementations of DFT intrinsically tend to underestimate the Coulomb repulsion and, hence, fail to capture the correlation-driven charge localizations. Consequently, the conventional DFT calculations appear to underestimate or even close the band gap. They may even result in an incorrect description of the magnetic ground state. 3 Among the remedies to overcome this problem, the so-called LDA+ U approach 3 has already proven to be computationally one of the most efficient methods to study a large variety of strongly correlated systems, with considerable improvements over LDA and GGA. In this method, a strong intra-atomic interaction is introduced in a screened Hartree-Fock-type manner as on-site replacement of the LDA ͑GGA͒. Since the introduction of LDA+ U by Anisimov et al., 3 this formalism has been extensively utilized to investigate the electronic and magnetic properties of narrow-band systems. However, it has rarely been used to predict the correct structural phase stability of strongly correlated systems.It is already known that for some compounds, both LDA and GGA severely fail to predict the ground-state structure. An example for such systems is CeMnNi 4 . It is a unique intermetallic soft ferromagnet ͑FM͒ discovered recently, 4 which exhibits large magnetic moment ͑ϳ4.95 B / Mn͒, reasonably high Curie temperature ͑ϳ150 K͒, and a high degree of ...