{"data":{"kind":"file","path":"README.md","version_id":"xut7f1e8jrdfn4qd9gt15y24","entry":{"name":"README.md","path":"README.md","is_directory":false,"size":1679,"modified_at":"2025-08-31T03:32:46.855000","content_hash":"ac01d466c95f04748777f80e863cec6099b3b24505df8e3935565b863d55c798"},"entries":[],"content":"# toxicity-explanation\n\n### Overview\n- **Environment ID**: `toxicity-explanation`\n- **Short description**: Judge-based evaluation for toxicity classification with explanations using Civil Comments.\n- **Tags**: toxicity, classification, explanation, judge, single-turn\n\n### Datasets\n- **Primary dataset(s)**: `google/civil_comments` mapped to toxicity targets and metadata\n- **Source links**: Hugging Face Datasets\n- **Split sizes**: Train split; size optionally limited via `max_examples`\n\n### Task\n- **Type**: single-turn\n- **Parser**: default `Parser`\n- **Rubric overview**: `JudgeRubric` with a numeric (0–10) rubric normalized to 0–1; evaluates correctness and explanation quality\n\n### Quickstart\nRun an evaluation with default settings:\n\n```bash\nuv run vf-eval toxicity-explanation\n```\n\nConfigure model and sampling:\n\n```bash\nuv run vf-eval toxicity-explanation \\\n  -m gpt-4.1-mini \\\n  -n 20 -r 3 -t 1024 -T 0.7 \\\n  -a '{\"judge_model\": \"gpt-4.1-mini\", \"judge_base_url\": \"https://api.openai.com/v1\", \"judge_api_key_var\": \"OPENAI_API_KEY\", \"max_examples\": -1}'\n```\n\nNotes:\n- Use `-a` / `--env-args` to configure the judge model/provider and dataset size.\n\n### Environment Arguments\n| Arg | Type | Default | Description |\n| --- | ---- | ------- | ----------- |\n| `judge_model` | str | `\"gpt-4.1-mini\"` | Judge model name |\n| `judge_base_url` | str | `\"https://api.openai.com/v1\"` | Judge provider base URL |\n| `judge_api_key_var` | str | `\"OPENAI_API_KEY\"` | Env var containing judge API key |\n| `max_examples` | int | `-1` | If > 0, limit dataset to this many examples |\n\n### Metrics\n| Metric | Meaning |\n| ------ | ------- |\n| `reward` | Normalized judge score (0–1) |\n","encoding":"utf-8","truncated":false,"total_bytes":1679},"status":null}