This repo contains the artifacts produced from the experiments reported in the paper "Computationally Efficient Active Learning for Large Imbalanced Datasets" (Lesci and Vlachos, 2024).

Each subfolder contains a README.md file describing its contents. In summary:

- `main/` contains the main results for all datasets considered created using `bert-base-uncased`

- `other_models/` contains the results for two datasets using five additional models

- `ablations/` contains the outputs of the ablations for the proposed AnchorAL method