{"data":{"kind":"file","path":"README.md","version_id":"p9n1ezrlwr73hvto7dr4ntgz","entry":{"name":"README.md","path":"README.md","is_directory":false,"size":1553,"modified_at":"2025-09-14T21:32:01.907000","content_hash":"46e54b1a3d9965a2a474335d83b958d2078e6428b3dad05b36cb409a5549b187"},"entries":[],"content":"# mancala\n\n### Overview\n- **Environment ID**: `mancala`\n- **Short description**: Multi-turn Mancala game with strategic stone movement and capture mechanics\n- **Tags**: games, multi-turn, strategy, mancala, xml, feedback\n\n### Datasets\n- **Primary dataset(s)**: Synthetic Mancala game scenarios with varying difficulty levels\n- **Source links**: Generated programmatically\n- **Split sizes**: 1000 training examples, 100 evaluation examples (configurable)\n\n### Task\n- **Type**: multi-turn (game interaction)\n- **Parser**: `XMLParser` with `think`/`move` fields\n- **Rubric overview**: Win reward, efficiency reward, capture reward, and format compliance\n\n### Quickstart\nRun an evaluation with default settings:\n\n```bash\nuv run vf-eval mancala\n```\n\nConfigure model and sampling:\n\n```bash\nuv run vf-eval mancala \\\n  -m gpt-4.1-mini \\\n  -n 20 -r 3 -t 1024 -T 0.7 \\\n  -a '{\"num_train_examples\": 1000, \"num_eval_examples\": 100}'\n```\n\nNotes:\n- Use `-a` / `--env-args` to pass environment-specific configuration as a JSON object.\n\n### Environment Arguments\n\n| Arg | Type | Default | Description |\n| --- | ---- | ------- | ----------- |\n| `num_train_examples` | int | `1000` | Number of training examples |\n| `num_eval_examples` | int | `100` | Number of evaluation examples |\n\n### Metrics\n\n| Metric | Meaning |\n| ------ | ------- |\n| `win_reward` | 1.0 if Player 0 wins, 0.0 otherwise |\n| `efficiency_reward` | Higher score for fewer moves to win |\n| `capture_reward` | Bonus for strategic stone captures |\n| `format_reward` | Proper XML formatting compliance |\n\n","encoding":"utf-8","truncated":false,"total_bytes":1553},"status":null}