AGENTDARS/generated-critiques

---
license: cc-by-4.0
---
## Dataset Summary
This dataset presents critiques for 43k patches from 1,077 open-source python projects. The patches are generated by [Nebius's agent](https://huggingface.co/datasets/nebius/SWE-agent-trajectories) and the critiques are generated by GPT-4o.

How to use:
```python
from datasets import load_dataset
dataset = load_dataset("AGENTDARS/generated-critiques")
```

## Software Patch Evaluation Dataset

This dataset contains structured information about software patches, problem statements, evaluations, and critiques. It is designed to assess and compare different patches proposed for GitHub issues.

### Dataset Fields

The dataset consists of several key fields:

| Field Name | Description |
|--------------------|-------------|
| `instance_id` | A unique identifier assigned to each data instance. |
| `generated_patch` | The code changes proposed as a fix for the given issue. |
| `patch` | The actual patch being tested, containing modifications to the codebase. |
| `problem_statement` | A textual description of the issue that the patch is attempting to resolve. |
| `FAIL_TO_PASS` | Test cases that fail before applying the patch but should pass after the patch application. |
| `PASS_TO_PASS` | Test cases that pass both before and after applying the patch. |
| `p2p_failed` | Test cases from `PASS_TO_PASS` that fail after applying the generated patch. |
| `f2p_failed` | Test cases from `FAIL_TO_PASS` that still fail after applying the generated patch. |
| `prompt` | The prompt used to generate the critique. |
| `critique` | The critique generated by GPT-4o. |
| `critique_gt` | The critique for the ground truth patch. |

## License
The dataset is licensed under the Creative Commons Attribution 4.0 license. However, please respect the license of each specific repository on which a particular instance is based.