Datasets:

All datasets and evaluation Results

• All datasets and Evaluation Results are available in the Google Drive. To access the datasets and results: Click Here

Raw Dataset Sheet links:

• GIT RAW LINK SCRAPED DATASET: Click Here

• TSDetect Labeled Dataset: Click Here

• TsDetect + Gemini Natural Language Labeled Dataset: Click Here

Dataset Published in Hugging face for Fine Tuning LLM for Test Smell Detection

• TS Detect + Gemini Explained Dataset: Click Here

• Human Labeled Dataset for only 4 test smells: Click Here

Evaluations:

• Evaluation Results: Click Here

Fine Tuned Model

Our Fine Tuned Model is available at Hugging Face Model Hub. You can use it for your own test smell detection task. Click here to access the model: Click Here

Code Links:

Code extractor script from GitHub

Click Here

Raw file extractor script

Click Here

Default Branch Mapper

Click Here

Test Code Parser From Git Raw Link CSV:

Click Here

Initial Test Smell Cleaning and Preprocessing:

Click Here

Test Code - Production Code Downloading Script:

Click Here

CSV Formatter for TsDetect:

Click Here

LLM Friendly Dataset Generator:

Helpers:

• CSV Splitter
• CSV Combiner

LLM Friendly Dataset Generator:

Click here

Fine Tuning Notebooks:

1. Fine tuning with only labeled data for four test smell using unsloth: click here

2. Fine tuning with tsDetect Generated gemini explained dataset using unsloth: click here

3. Raw Fine tuning with tsDetect Generated gemini explained dataset: click here

Evaluation:

Helpers:

• Groq Agent
• huggingFace Agent
• HuggingFace Middleware (Serializer)
• Manual Model Loader
• Finetuned Model Evaluator
• TP, FP, TN, FN Calculator

GPT & Claude Agent:

Click Here

Overall Evaluator:

Click Here

Evaluation Results:

Click Here

Mixtral Of Agent(MOA) Architecture:

Click Here