Calibration-robust entanglement detection beyond Bell inequalities

Most entanglement verification is examined in either the completely characterized or the totally device-independent scenario. The assumptions imposed by these extreme cases are often either too weak or too strong for real experiments. Here we investigate this detection task in the intermediate regime where there is partial knowledge of the measured observables, considering cases like orthogonal, sharp, or only dimension-bounded measurements. We show that for all these assumptions it is not necessary to violate a corresponding Bell inequality in order to detect entanglement. We derive strong detection criteria that can be directly evaluated for experimental data and which are robust against large classes of calibration errors. The conditions are even capable of detecting bound entanglement under the sole assumption of dimension-bounded measurements.