Update README.md

README.md

@@ -15,6 +15,10 @@ The Recode benchmark proposes to apply code and natural language transformations
 This dataset contains the perturbed version of HumanEval that they released.
 It was automatically generated from the [HumanEval](https://huggingface.co/datasets/openai_humaneval) dataset.
 
+### Subsets
+
+There are four transformation categories that form the subsets of this dataset: `func_name`, `nlaugmenter`, `natgen` and `format`.
+
 ### Languages
 
 The programming problems are written in Python and contains docstrings and comments in English.
@@ -27,11 +31,19 @@ The programming problems are written in Python and contains docstrings and comme
 
 ### Data Fields
 
-[More Information Needed]
+- `task_id`: ID of the original HumanEval example
+- `prompt`: the perturbed prompt
+- `entry_point`: entry point for test
+- `canonical_solution`: solution for the problem in the `prompt`
+- `test`: contains function to test generated code for correctness
+- `seed`: seed of the perturbed prompt
+- `perturbation_name`: name of the perturbation
+- `partial`: partial solution to the problem. This field is only present for transformation categories that affect a partial solution: `natgen` and `format`.
+
 
 ### Data Splits
 
-[More Information Needed]
+The dataset only has a test split.
 
 ## Dataset Creation
 
@@ -89,7 +101,12 @@ The programming problems are written in Python and contains docstrings and comme
 
 ### Citation Information
 
-[More Information Needed]
+@article{wang2022recode,
+  title={ReCode: Robustness Evaluation of Code Generation Models},
+  author={Wang, Shiqi and Li, Zheng and Qian, Haifeng and Yang, Chenghao and Wang, Zijian and Shang, Mingyue and Kumar, Varun and Tan, Samson and Ray, Baishakhi and Bhatia, Parminder and others},
+  journal={arXiv preprint arXiv:2212.10264},
+  year={2022}
+}
 
 ### Contributions