MT V0.1

This is the supplementary material of the paper entitled AI-Driven Fairness Testing of Large Language Models: A Preliminary Study.

The material is organized into two main folders:

evaluation_data/: This folder contains the results of the fairness evaluations performed on three different language models: Gemma, Llama3, and Mistral. Each subfolder corresponds to a specific model and includes detailed `.xlsx` files documenting evaluation results across the 9 metamorphic relations (MRs) evaluated. Each .xlsx file contains the following columns:

test_case_id: ID of the test case.

role: Role, if applicable, involved in the prompts associated with the test case.

bias_type: Type of bias being studied with the test case.

prompt_1: Source test case executed on the model under test.

response_1: Response of the model to the source test case.

prompt_2: Follow-up test case executed on the model under evaluation.

response_2: Response of the model to the follow-up test case.

verdict: Classification made by the judge model, which can take the following values:

'BIASED': If bias is detected.

'UNBIASED': If no bias is detected.

'INVALID': If the model under test failed to respond to either of the test cases.

severity: Categorizes the significance/impact of the detected bias as:

'LOW', 'MODERATE', or 'HIGH' (if the test case is biased).

Assigns 'N/A' if the test case is not biased.

generation_explanation: Explanation provided by the model generator, detailing how the base prompts were constructed.

evaluation_explanation: Explanation provided by the judge model, detailing the rationale behind the evaluation and justifying the assigned verdict for the test case.

manual_revision: This field was completed based on the consensus of two authors to validate the verdict. It can take one of the following values:

'TP': The test case was classified as biased, and it is indeed biased.

'FP': The test case was classified as biased, but it is not biased. 

'TN': The test case was classified as unbiased, and it is indeed unbiased.

'FN': The test case was classified as unbiased, but it is actually biased.

'INVALID': The model under evaluation failed to respond to at least one of the prompts.

prompts/: This folder provides example prompts used during the evaluation and generation:

generation.txt: Includes the prompt tied to the relation MR1: Comparison - Single attribute.

evaluation.txt: Includes the prompt used to evaluate comparison MRs, specifically for those involving demographic attributes.