{"data":{"kind":"file","path":"README.md","version_id":"kjjydrezcps6nyqetcsudbka","entry":{"name":"README.md","path":"README.md","is_directory":false,"size":1142,"modified_at":"2026-05-22T07:51:22.736000","content_hash":"5b60fb112c8feddd8bfd50101dd55fa0c96cf4a05b6ee14f0d7a84ee657a9cae"},"entries":[],"content":"# reverse-text\n\nSingle-turn environment that asks the model to reverse a string character by\ncharacter. Partial credit comes from a longest-common-subsequence ratio against\nthe true reversal.\n\n### Overview\n- **Environment ID**: `reverse-text`\n- **Short description**: Reverse input text character-by-character with tagged output.\n- **Tags**: single-turn, text, train, eval\n\n### Datasets\n- **Primary dataset(s)**: [PrimeIntellect/Reverse-Text-RL](https://huggingface.co/datasets/PrimeIntellect/Reverse-Text-RL) train split\n- **Split sizes**: full train split for task rows\n\n### Task\n- **Type**: single-turn\n- **Output format expectations**: answer inside `<reversed_text>...</reversed_text>` tags\n- **Scoring**: LCS ratio between parsed answer and reversed input\n\n### Quickstart\n\n```bash\nprime env install reverse-text -p ./environments\nprime eval run reverse-text\n```\n\nConfigure model and sampling:\n\n```bash\nprime eval run reverse-text \\\n  -m openai/gpt-4.1-mini \\\n  -n 20 -r 3 -t 1024 -T 0.7\n```\n\n### Metrics\n\n| Metric | Meaning |\n| ------ | ------- |\n| `lcs_reward` | LCS ratio between parsed `<reversed_text>` answer and target reversal |\n","encoding":"utf-8","truncated":false,"total_bytes":1142},"status":null}