FuseGen

This is the code for paper FuseGen.

We implement our code for data generation basing on the code provided in this repo.

Key Components:

1. Cross-model Data Quality Evaluation in Cross-model Data Generation:
   1. Cross-model Variability: Please see function sample_dynamic_selection in file ./src/utils/basic_utils.py.
   2. Sample Influence: Please see function run_full_influence_functions in file ./src/utils/influence_utils.py.
2. Cross-model Data Quality Improvement:
   1. Self-boosting Weight Adjustment: Please see function weight_decay in ./src/utils/weight_adjust.py.