{"data":{"kind":"file","path":"README.md","version_id":"i44a6xel5g5s2k82vvnmfpzu","entry":{"name":"README.md","path":"README.md","is_directory":false,"size":1285,"modified_at":"2026-04-09T16:38:02.197000","content_hash":"f4ab7e004bba95161d7c429920b0f88b4465b6d2d6e53ab6cb7fb58eb8020d39"},"entries":[],"content":"# unhinged\n\n### Overview\n- **Environment ID**: `ziliangpeng/unhinged`\n- **Short description**: Train a model to speak casually with natural profanity — angry and unhinged no matter the input\n- **Tags**: chat, personality, train, eval\n\n### Datasets\n- **Primary dataset(s)**: 192 built-in prompts (128 frustrating scenarios + 64 normal/friendly conversations)\n- **Split sizes**: All prompts used for training, repeated to fill `num_examples`\n\n### Task\n- **Type**: single-turn\n- **Rubric overview**: Rule-based reward scoring swear word count, ALL CAPS rage, exclamation marks, rage emojis. Penalizes corporate speak, euphemisms, and hedging language.\n\n### Quickstart\nRun an evaluation with default settings:\n\n```bash\nprime eval run ziliangpeng/unhinged\n```\n\nConfigure model and sampling:\n\n```bash\nprime eval run ziliangpeng/unhinged -m gpt-4.1-mini -n 20 -r 3 -t 1024 -T 0.7\n```\n\n### Environment Arguments\n\n| Arg | Type | Default | Description |\n| --- | ---- | ------- | ----------- |\n| `num_examples` | int | `500` | Total dataset size (prompts are repeated to fill) |\n\n### Metrics\n\n| Metric | Meaning |\n| ------ | ------- |\n| `reward` | Weighted sum: swear words (up to 0.5) + caps bonus (0.1) + exclamation bonus (0.1) + emoji bonus (0.15) - corporate/euphemism/hedging penalties |\n","encoding":"utf-8","truncated":false,"total_bytes":1285},"status":null}