CortexLM/swe-forge

---
task_categories: ["text-generation"]
license: apache-2.0
tags: ["swe-bench", "code-generation", "software-engineering", "benchmark"]
size_categories: ["n<1K"]
---

# SWE-Forge Dataset

**20 validated tasks** for evaluating software engineering agents.

Each task contains:
- `workspace.yaml` - Task configuration (repo, commits, install commands, test commands)
- `patch.diff` - The ground-truth patch
- `tests/` - Generated test files (fail before patch, pass after)
- `evaluate.sh` - Binary evaluator (score 0 or 1)

## Docker Images

Pre-built images on Docker Hub: `platformnetwork/swe-forge:<task_id>`

Each image has the repo cloned at base_commit with dependencies installed.
The benchmark runner applies the agent's patch then mounts tests to evaluate.

## Usage

```python
from datasets import load_dataset
ds = load_dataset("CortexLM/swe-forge")
```