Background: Odd-odd nuclei, around doubly closed shells, have been extensively used to study proton-neutron interactions. However, the evolution of these interactions as a function of the binding energy, ultimately when nuclei become unbound, is poorly known. The 26 F nucleus, composed of a deeply bound π 0d 5/2 proton and an unbound ν0d 3/2 neutron on top of an 24 O core, is particularly adapted for this purpose. The coupling of this proton and neutron results in a J π = 1 1 + − 4 1 + multiplet, whose energies must be determined to study the influence of the proximity of the continuum on the corresponding proton-neutron interaction. The J π = 1 1 + , 2 1 + , 4 1 + bound states have been determined, and only a clear identification of the J π = 3 1 + is missing. Purpose: We wish to complete the study of the J π = 1 1 + − 4 1 + multiplet in 26 F, by studying the energy and width of the J π = 3 1 + unbound state. The method was first validated by the study of unbound states in 25 F, for which resonances were already observed in a previous experiment. Method: Radioactive beams of 26 Ne and 27 Ne, produced at about 440A MeV by the fragment separator at the GSI facility were used to populate unbound states in 25 F and 26 F via one-proton knockout reactions on a CH 2 target, located at the object focal point of the R 3 B/LAND setup. The detection of emitted γ rays and neutrons, added to the reconstruction of the momentum vector of the A − 1 nuclei, allowed the determination of the energy of three unbound states in 25 F and two in 26 F. Results: Based on its width and decay properties, the first unbound state in 25 F, at the relative energy of 49(9) keV, is proposed to be a J π = 1/2 − arising from a p 1/2 proton-hole state. In 26 F, the first resonance at 323(33) keV is proposed to be the J π = 3 1 + member of the J π = 1 1 + − 4 1 + multiplet. Energies of observed states in 25,26 F have been compared to calculations using the independent-particle shell model, a phenomenological shell model, and the ab initio valence-space in-medium similarity renormalization group method. Conclusions: The deduced effective proton-neutron interaction is weakened by about 30-40% in comparison to the models, pointing to the need for implementing the role of the continuum in theoretical descriptions or to a wrong determination of the atomic mass of 26 F.