{"data":{"kind":"file","path":"README.md","version_id":"xfxnlej4yz2bph5nd7oi8pd8","entry":{"name":"README.md","path":"README.md","is_directory":false,"size":1713,"modified_at":"2026-06-16T20:01:05.787000","content_hash":"1f3791800055c10f7a3d73b59fc7be0e4ce2cf69edfc88c9b1d9620c333bd7c1"},"entries":[],"content":"# meta-writing-style-detector\n\n`meta-writing-style-detector` is a deterministic Verifiers environment for\nprobing whether small models can learn a lightweight theory of writing style.\nThe task is binary authorship/style detection: decide whether a sentence is\nderived from Lewis Carroll's public-domain *Alice's Adventures in Wonderland*\nor is a generated imitation about similar characters and situations.\n\nThe environment is deliberately cheap:\n\n- no tools\n- no sandbox\n- no judge model\n- deterministic source, imitation, and counterfeit generation\n- compact JSON output\n\nPrompts ask for exactly one result tag:\n\n```text\n<result>{\"label\":\"book\",\"confidence\":0.82}</result>\n```\n\nAllowed labels are `book` and `imitation`. Confidence must be a number from\n`0.0` to `1.0`.\n\nMetrics separate classification from formatting and calibration:\n\n- `label_exact`\n- `book_correct`\n- `imitation_correct`\n- `predicted_book`\n- `parseable`\n- `exact_one_result`\n- `schema_valid`\n- `confidence_in_range`\n- `calibration_score`\n- `overconfident_wrong`\n- `underconfident_correct`\n- `raw_json`\n- `output_chars`\n\nSource text notes:\n\n- Book sentences are manually selected and ASCII-normalized from the Project\n  Gutenberg text of *Alice's Adventures in Wonderland*.\n- `easy`, `balanced`, and `near_miss` imitation sentences are deterministic\n  local templates, not model-generated.\n- `counterfeit` imitation sentences are harder negatives made by lightly\n  altering or splicing real book sentences, so they preserve much more of the\n  source style while no longer being exact copied text.\n- Future versions can replace the imitation generator with recorded LLM\n  generations, while keeping the same parser, metrics, and eval matrix.\n","encoding":"utf-8","truncated":false,"total_bytes":1713},"status":null}