This dataset contains reasoning traces generated by the QwQ system applied to the Knights and Knaves logic puzzles. Each example in the dataset includes a natural language puzzle, the reasoning steps taken by the model, and the tokens used. These traces reflect intermediate reasoning steps that help interpret how QwQ reaches a conclusion.