Explicit and implicit conflicts of floating point rounding error analysis in Financial interest computation

This dataset contains 1.85M ground-truth annotated conflict cases (355k explicit, 1.49M implicit) for high-precision Decimal floating-point rounding error detection, discovered using Cascaded DWT-Newton framework without historical data, validated against Decimal prec=1000 reference, and includes 3,902 explicit conflict arbitration samples with 12-dimensional perturbation-based sensitivity features for training machine learning models.