README.md

---
dataset_info:
  features:
  - name: file_name
    dtype: string
  - name: verified_c_file_content
    dtype: string
  - name: dependency_acls
    dtype: string
  - name: function_implementation
    dtype: string
  - name: verified
    dtype: bool
  - name: total_goals
    dtype: int64
  - name: verified_goals
    dtype: int64
  - name: error_cause
    dtype: string
  - name: c_code_snippet
    dtype: string
  - name: acsl_snippet
    dtype: string
  splits:
  - name: train
    num_bytes: 1702402
    num_examples: 506
  download_size: 444955
  dataset_size: 1702402
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

## General information:
This dataset is meant for evaluating the ability of LLMs to generate a valid implementation given a specification and is partly a derivative of The Stack 1 and The Stack 2. 
The dataset is meant for research purposes. Please look up the legal license from The Stack 1 and The Stack 2 before using.  

## Please cite our paper if you use this dataset:
Here is a link to our paper describing the process of creating this dataset: https://link.springer.com/chapter/10.1007/978-3-032-07132-3_5    

## Example of use:
The following results were obtained by prompting the LLMs to generate an implementation given an ACSL specification and the function signature. We only used 100 examples for this test. The outputs were then evaluated using Frama-C with the WP and RTE plugins, for more information of specific versions etc, please have a look at our paper.  
![model_comparison_stacked_ranked](https://cdn-uploads.huggingface.co/production/uploads/63a06a23f3334a6553d9d5d1/LmwBD6q6yPaF37dKdgaIS.png)

## Contact
If you have any questions regarding the dataset, you can email me at niclas.hertzberg@ai.se.