{"data":{"kind":"file","path":"README.md","version_id":"b1bdf1a1r4o0jahc49wcqn5r","entry":{"name":"README.md","path":"README.md","is_directory":false,"size":2672,"modified_at":"2026-07-09T09:54:12.817000","content_hash":"1e193139581b7ebf6277f77b5366b7a8b860fc91e8afc4216c742ff0fe4b89e2"},"entries":[],"content":"# irreps-tensor-product\n\n<a href=\"https://github.com/PrimeIntellect-ai/verifiers/tree/main/environments/irreps_tensor_product\">\n<img src=\"https://img.shields.io/badge/GitHub-181717?style=for-the-badge&logo=github&logoColor=white\" alt=\"Source Code\">\n</a>\n\n### Overview\n- **Environment ID**: `irreps-tensor-product`\n- **Short description**: Single-turn O(3) irreducible representation tensor product decompositions with exact symbolic scoring.\n- **Tags**: math, representation-theory, single-turn, eval\n\n### Datasets\n- **Primary dataset(s)**: Procedurally generated by this environment\n- **Source links**: local deterministic generator\n- **Split sizes**: Configurable counts for train/eval via loader args\n\n### Task\n- **Type**: single-turn\n- **Rubric overview**: Exact canonical symbolic match after whitespace normalization.\n- **Prompt contract**: The model receives the tensor-product rule in the prompt.\n- **Answer contract**: Canonical decompositions use ascending e3nn-style irreps separated by ` + `, such as `1o + 2o + 3o`.\n\n### Quickstart\nRun an evaluation with default settings:\n\n```bash\nprime eval run irreps-tensor-product\n```\n\nConfigure model, sampling, and task difficulty:\n\n```bash\nprime eval run irreps-tensor-product \\\n  -m openai/gpt-4.1-mini \\\n  -n 20 -r 3 -t 1024 -T 0.7 \\\n  -a '{\"num_train_examples\": 2000, \"num_eval_examples\": 2000, \"seed\": 0, \"min_l\": 0, \"max_l\": 4}'\n```\n\nNotes:\n- Use `-a` / `--env-args` to pass environment-specific configuration as a JSON object.\n- The Python module name is `irreps_tensor_product`; the CLI environment ID uses `irreps-tensor-product`.\n- The task is standalone and does not require `reasoning-gym`.\n- e3nn notation is used for task notation only. e3nn is not a runtime dependency of this environment.\n\n### Environment Arguments\n| Arg | Type | Default | Description |\n| --- | ---- | ------- | ----------- |\n| `num_train_examples` | int | `2000` | Number of training examples |\n| `num_eval_examples` | int | `2000` | Number of evaluation examples |\n| `seed` | int | `0` | Random seed for dataset generation |\n| `min_l` | int \\| null | `null` | Optional minimum angular momentum degree |\n| `max_l` | int \\| null | `null` | Optional maximum angular momentum degree |\n| `include_even` | bool \\| null | `null` | Optional pass-through parity control |\n| `include_odd` | bool \\| null | `null` | Optional pass-through parity control |\n| `system_prompt` | str \\| null | `null` | Optional Verifiers system prompt override |\n\n### Metrics\n| Metric | Meaning |\n| ------ | ------- |\n| `reward` | Exact score for the parsed canonical decomposition |\n| `format_reward` | Adherence to the XML answer format; currently weight `0.0` |\n","encoding":"utf-8","truncated":false,"total_bytes":2672},"status":null}