Multi-task profile model performance across all 10 folds

Single-task profile model performance across all 10 folds

Fine-tuned multi-task profile model task-specific performance

Multi-task profile model performance across all 10 folds ¶

A comparison of the test-set performance (averaged across tasks) between:

  1. Multi-task profile models trained across all 10 folds
  2. A fine-tuned multi-task profile model on the best-performing fold
  3. Upper and lower bounds on the best-performing fold

Single-task profile model performance across all 10 folds ¶

For each task , a comparison of the test-set performance between:

  1. Single-task profile models trained across all 10 folds
  2. A fine-tuned single-task profile model on the best-performing fold
  3. Upper and lower bounds on the best-performing fold

Task 0

Task 1

Task 2

Task 3

Fine-tuned multi-task profile model task-specific performance ¶

A comparison of the test-set performance for between each task of a multi-task profile model fine-tuned on the best-performing fold