How Many and Which Training Points Would Need to be Removed to Flip this Prediction?

Official implementation for the paper " How Many and Which Training Points Would Need to be Removed to Flip this Prediction?"

Instruction

1. The Algorithm 1 is in IP function under Smallest_k.py. The Algorithm 2 is in recursive_NT function under iterative.py.
2. To run algorithms for the SST dataset, you can use the command below:

mkdir results
python SST.py

3. To run algorithms for the SST feature extracted from BERT, you can use the command:

python SST_bert.py

4. You can download other processed datasets and hyperparameter information via https://huggingface.co/datasets/Eciel/Smallest_k_experiment.