Deleting uncertain observations from every dataset.

Making all variables strictly binary.

Reasoning with gradual degrees of truth rather than only true/false values.

Guaranteeing zero model error.

Silhouette score

RMSE

Accuracy

Learning rate

Dataset name and source, URL, short description, and license/usage conditions if available.

Only the file size, without source or description.

Only screenshots of dataset rows.

No dataset information, because models are more important.

Replacing all missing values with labels.

Visualizing high-dimensional data in two or three dimensions.

Guaranteeing production-level model fairness.

Optimizing database transactions.

Represent words as dense numerical vectors that capture semantic relationships.

Remove all stop words without replacement.

Store documents only as image files.

Convert models into hardware devices.

A justified comparison showing how performance or efficiency changed relative to the baseline.

Only a screenshot of code without results.

A comparison using a different dataset and different metric without explanation.

Only a statement that the new model is modern.

A rule that always predicts the majority class.

A decision boundary that maximizes the margin between classes.

A set of random clusters with no target labels.

A database schema for storing images.

Only raw code with no explanation.

Only a model name without metrics.

Tables of metrics, plots such as confusion matrices or learning curves, and interpretation of what worked best and why.

Only a conclusion written before experiments are performed.

Apriori

t-SNE

ARIMA

LSTM

Creating interactive web dashboards only.

Automatically labeling unstructured datasets.

Replacing all model evaluation metrics.

Efficient numerical operations with arrays and matrices.