This is part 8/10 of the system predictions for the Real-Bug Test set. This system achieves 41% Repair Accuracy and 54% Location Accuracy.
Annotations:
- foo → bar - The bug was located at this node, and the system correctly predicted both the location and the repair as its top prediction.
- foo → bar - The system incorrectly predicted this repair as its top prediction.
- foo → bar - If the system incorrectly predicted the repair, this was the correct repair it should have made.
- foo → bar, pred: zap - The system predicted the correct location of the repair, but the wrong repair label. It should have predicted 'bar', but it predicted 'zap' instead.
The part before the arrow ('foo') is what the system actually saw at test time. Other candidate repair locations which the system could have chosen are marked in this color. For clarity the actual labels for those locations are not shown.