AsserT5: Test Assertion Generation Using a Fine-Tuned Code Language Model (Replication Package)

Replication package for our AST 2025 paper ‘AsserT5: Test Assertion Generation Using a Fine-Tuned Code Language Model’ (preprint: https://arxiv.org/abs/2502.02708).

Contents of the replication package:

data/: our extended methods2test training & test dataset

models/: our implementations of the models

results/: raw results and processed ones using the evaluation scripts

tools/: other tools and scripts we used during evaluation

AssertionCommander (model query & metrics evaluation) also available at https://github.com/sprimbs8-up/AssertionCommander 

AssertionExtractor (extract assertion statements from test cases) also available at https://github.com/sprimbs8-up/AssertionExtractor

methods2testcrawler also available at https://github.com/sprimbs8-up/methods2test-crawler

checkpoints/: The model checkpoints for the trained AsserT5 variants and the comparison models.

Only the checkpoints for the best-performing model variant with up to 10 assertions per test case are included.

The other variants can be found on the Hugging Face Hub at https://huggingface.co/se2p/AsserT5 (https://doi.org/10.57967/hf/4225)