AhmedTaha012/AUTOSAR_CONFIGURATION_PROMPT_TO_JSON_V2

AUTOSAR Configuration Prompt to JSON

Dataset Description

The AUTOSAR Configuration Prompt to JSON dataset is designed for developing and evaluating machine learning and large language models that convert natural-language prompts describing AUTOSAR configuration requirements into structured JSON representations.

The dataset supports experimentation with prompt understanding, structured output generation, and the transformation of automotive software configuration requirements into machine-readable formats.

Intended Use

Potential applications include:

• Fine-tuning language models for AUTOSAR configuration tasks.
• Training models to transform natural-language requirements into JSON.
• Evaluating structured JSON generation and instruction-following capabilities.
• Experimenting with supervised fine-tuning (SFT) and prompt-based approaches.
• Building AI assistants for automotive software configuration workflows.

Dataset Structure

The dataset consists of three CSV files corresponding to the training, validation, and test splits.

File	Split	Purpose
Train.csv	train	Used to train and fine-tune models.
Vald.csv	validation	Used for validation, model selection, and hyperparameter tuning.
Test.csv	test	Reserved for final evaluation on held-out examples.

Data Format

All files are provided in CSV format.

Each row represents a dataset example. The exact columns, input prompt format, target JSON representation, and any additional metadata depend on the source files.

Inspect the CSV headers to determine which columns represent the input prompts and expected outputs.

Tasks

The dataset is intended to support the following tasks:

1. Prompt Understanding: Interpret natural-language AUTOSAR configuration requests.
2. Structured Output Generation: Generate JSON objects from configuration prompts.
3. Instruction Following: Produce outputs that follow the requested structure and constraints.
4. Model Evaluation: Measure the correctness, completeness, and consistency of generated JSON.
5. Domain-Specific Language Modeling: Explore language model performance on automotive configuration scenarios.

Recommended Evaluation

Recommended evaluation methods include:

• JSON Parse Rate: Percentage of generated outputs that are syntactically valid JSON.
• Exact Match: Percentage of predictions that exactly match the expected output.
• Structural Accuracy: Correctness of JSON keys, values, types, and nesting.
• Field-Level Accuracy: Accuracy of individual configuration fields.
• Schema Validation: Compliance with the relevant JSON schema or AUTOSAR configuration rules, where applicable.

These are suggested evaluation methods; no benchmark results are claimed by this dataset card.

Recommended Usage

Use the Train.csv split for model training, Vald.csv for validation and model selection, and Test.csv for final evaluation.

Avoid using test examples during training or hyperparameter tuning to prevent evaluation leakage.

Generated JSON should be validated against the appropriate schemas and configuration constraints before being used in an engineering workflow.

Limitations

• Dataset coverage depends on the examples available in the source data.
• Natural-language prompts may be ambiguous or incomplete.
• A syntactically valid JSON output is not necessarily a semantically correct AUTOSAR configuration.
• Model performance may vary across AUTOSAR versions, configuration modules, and prompt styles.
• Generated outputs should not be assumed to be production-ready without validation.

Data Quality and Validation

Before training, users should verify:

• CSV encoding and column consistency across all three files.
• Missing values and malformed records.
• Duplicate or overlapping examples across splits.
• Correct alignment between input prompts and target JSON outputs.
• The validity of expected JSON outputs.

Intended Audience

This dataset is intended for researchers, machine learning engineers, AI developers, and automotive software engineers working on language models and structured configuration generation.

License

The license and redistribution permissions for the source data have not been specified in this dataset card. Please confirm the applicable rights before distributing the dataset publicly.

Citation

If you use this dataset in research or development, cite the Hugging Face repository:

AhmedTaha012/AUTOSAR_CONFIGURATION_PROMPT_TO_JSON

Please refer to the repository for the latest dataset version and documentation.