Frozen logits and known labels
Reference inputs · synthetic data
- Records
- 3
- Data columns
- 4
- Missing cells
- 0
- Duplicate records
- 0
- Feature duplicates
- 1
3 of 3 records · One record = one observation. Statistics and downloads use all records.
About this dataset
Three constructed examples with fixed model scores. Two have identical logits but different true labels, letting us investigate confidence independently of accuracy.
- A and B share the same features but have different labels. They are feature duplicates, not duplicate complete rows; do not automatically remove either.
- These three examples cannot establish population calibration or model generalization.
- Temperature and common offset transform these frozen scores during the experiment; the source logits remain unchanged.
Counts exclude the row identifier. Duplicate counts are extra occurrences after the first. Feature duplicates ignore labels and experimental conditions. — means a statistic is undefined or not applicable.
- Row
- Stable identifier within this synthetic fixture.
- Logit 0
- Frozen score for class 0; not a probability.
- Logit 1
- Frozen score for class 1; not a probability.
- Logit 2
- Frozen score for class 2; not a probability.
- True class
- Known class label, independent of the model’s score.