QUACQ is a constraint acquisition system that as- sists a non-expert user to model her problem as a constraint network by classifying (partial) exam- ples as positive or negative. For each negative ex- ample, QUACQ focuses onto a constraint of the tar- get network. The drawback is that the user may need to answer a great number of such examples to learn all the constraints. In this paper, we provide a new approach that is able to learn a maximum num- ber of constraints violated by a given negative ex- ample. Finally we give an experimental evaluation that shows that our approach improves on QUACQ.