1. Train and save model checkpoints ./less/scripts/train/finfluence_train_llama3.1_8b.sh 2. Run the FINE pipeline ./run_fine.sh 3. Compute LESS scores ./less_selection_pipeline.sh Notes Change paths and arguments in each shell file. Can run on a single RTX A6000.