Abstract model refinement for model-based program debugging

Automatic debugging of software has been an ac-tive research area for several years. We analyze an existing model-based approach for debugging and highlight its deficiencies regarding the reported faults. To overcome these deficiencies we present a new model that is extended with abstract infor-mation. The used abstraction is based on predicate abstraction, a method developed for abstract model checking of software. The abstract information in the model adapts itself as necessary in an automatic refinement process. Our new model produces bet-ter results than existing models as we show with a running example.