{"data":{"kind":"file","path":"README.md","version_id":"pc9moi5npt7wflpz3pgwn0d9","entry":{"name":"README.md","path":"README.md","is_directory":false,"size":2030,"modified_at":"2025-08-21T23:59:13","content_hash":"cdedf358da49248211c7036989bd7a4b02441905bd635a984393f20ab3d92587"},"entries":[],"content":"# self-reward\n\n### Overview\n- **Environment ID**: `self-reward`\n- **Short description**: Single-turn evaluation where a judge model scores responses based on a simple scoring prompt.\n- **Tags**: judge, single-turn, self-reward, openai-compatible\n\n### Datasets\n- **Primary dataset(s)**: Any HF dataset with `question`/`answer` columns (specified by `dataset_name`)\n- **Source links**: Hugging Face Datasets\n- **Split sizes**: Uses the dataset’s `train` file by default\n\n### Task\n- **Type**: single-turn\n- **Parser**: default `Parser`\n- **Rubric overview**: `JudgeRubric` uses a judge client/model/prompt to produce a 0–1 score\n\n### Quickstart\nRun an evaluation with default settings (example):\n\n```bash\nuv run vf-eval self-reward -a '{\"dataset_name\": \"your/dataset\", \"model_name\": \"gpt-4.1-mini\"}'\n```\n\nConfigure model and sampling:\n\n```bash\nuv run vf-eval self-reward \\\n  -m gpt-4.1-mini \\\n  -n 20 -r 3 -t 1024 -T 0.7 \\\n  -a '{\"dataset_name\": \"your/dataset\", \"model_name\": \"gpt-4.1-mini\", \"base_url\": \"http://0.0.0.0:8000/v1\", \"api_key_var\": \"JUDGE_API_KEY\"}'\n```\n\nNotes:\n- Use `-a` / `--env-args` to pass environment-specific configuration as a JSON object.\n- Reports are written under `./environments/vf_self_reward/reports/` and auto-embedded below.\n\n### Environment Arguments\n| Arg | Type | Default | Description |\n| --- | ---- | ------- | ----------- |\n| `dataset_name` | str | — | HF dataset name or path containing `question`/`answer` |\n| `model_name` | str | — | Judge model name (OpenAI-compatible) |\n| `base_url` | str | `\"http://0.0.0.0:8000/v1\"` | Judge API base URL |\n| `api_key_var` | str | `\"JUDGE_API_KEY\"` | Env var containing judge API key |\n\n### Metrics\n| Metric | Meaning |\n| ------ | ------- |\n| `reward` | Judge-produced score, normalized to 0–1 |\n\n## Evaluation Reports\n\n<!-- Do not edit below this line. Content is auto-generated. -->\n<!-- vf:begin:reports -->\n<p>No reports found. Run <code>uv run vf-eval vf-self-reward -a '{\"key\": \"value\"}'</code> to generate one.</p>\n<!-- vf:end:reports -->\n","encoding":"utf-8","truncated":false,"total_bytes":2030},"status":null}