{"data":{"kind":"file","path":"README.md","version_id":"nopeo8fp90hnh7jgkh0lxh65","entry":{"name":"README.md","path":"README.md","is_directory":false,"size":1583,"modified_at":"2026-02-05T20:15:11.299000","content_hash":"0193f36a0a7f89efc5ca61707c4eb1ab7bbb8d6f2e867ed28846d07b20b1b40d"},"entries":[],"content":"# animal-crossing-letter\n\n### Overview\n- **Environment ID**: `animal-crossing-letter`\n- **Short description**: Scores the points value of a response according to the algorithm used in Animal Crossing (GameCube, 2002)\n- **Tags**: <comma-separated tags>\n\n### Datasets\n- **Primary dataset(s)**: The algorithm was datamined by Hunter R, James Chambers, and Cuyler.\n- **Source links**: https://github.com/HunterRDev/AC-Letter-Scorer/tree/main\n\n### Task\n- **Type**: single-turn\n- **Parser**: The bare response is scored directly.\n- **Rubric overview**: Letters are evaluated on a heuristic which is based on punctuation placement and the presence of trigrams.\n\n### Quickstart\nRun an evaluation with default settings:\n\n```bash\nprime eval run animal-crossing-letter\n```\n\nConfigure model and sampling:\n\n```bash\nprime eval run animal-crossing-letter   -m gpt-4.1-mini   -n 20 -r 3 -t 1024 -T 0.7   -a '{\"key\": \"value\"}'  # env-specific args as JSON\n```\n\nNotes:\n- Use `-a` / `--env-args` to pass environment-specific configuration as a JSON object.\n\n### Metrics\nSummarize key metrics your rubric emits and how they’re interpreted.\n\n| Metric | Meaning |\n| ------ | ------- |\n| `reward` | Main scalar reward (weighted sum of criteria) |\n\n### Scoring Oracle And Fixtures\n\n- `letter.js` is the JS scoring oracle (bug-for-bug behavior source of truth).\n- `score_fixtures.json` stores frozen JS-generated expected outputs.\n- `test_score_fixtures.py` validates Python scoring against those fixtures.\n\nRegenerate fixtures with Bun:\n\n```bash\ncd animal_crossing_letter\nbun run generate_fixtures.js\n```\n","encoding":"utf-8","truncated":false,"total_bytes":1583},"status":null}