Revised evaluation artifact for the Modelizer framework described in "Learning Program Behavioral Models from Synthesized Input-Output Pairs"

This package contains the revised evaluation artifact for Modelizer - a framework for learning models from BlackBox systems using Input-Output examples. It consists of the following zip archives: train.zip: synthetic data used for the model training test.zip: synthetic and extracted real data used for the model evaluation eval.zip: evaluation results (experiment results and notebooks with the analysis of the results) eval_llm.zip: evaluation results for experiments with LLMs models.zip: all the models that were trained during the experiments llm_fine_tuned_models.zip: weights of fine-tuned the LLMs llm.zip: subset of training data for experiments with LLMs plots.zip: plots of the evaluation results and code to generate them You can find the framework implementation and usage instructions at https://github.com/2ral/Modelizer