{"data":{"kind":"file","path":"README.md","version_id":"hnqz1fy7n1hsckmas63feksj","entry":{"name":"README.md","path":"README.md","is_directory":false,"size":2123,"modified_at":"2025-09-07T14:36:45.928000","content_hash":"c657ef9f72e11d7f1e27d3aa30ef015a572856bc0b8a4a7e9558f18f53fd7f94"},"entries":[],"content":"# hendrycksmath\n\n### Overview\n- **Environment ID**: `hendrycksmath`\n- **Short description**: Single-turn hendrycksmath problems with boxed numeric answers and CoT.\n- **Tags**: math, single-turn, think, boxed-answer\n\n### Datasets\n- **Primary dataset(s)**: `EleutherAI/hendrycks_math` via `load_dataset(\"EleutherAI/hendrycks_math\")` (maps problem→question & solution→answer)\n- **Source links**: [EleutherAI/hendrycks_math](https://huggingface.co/datasets/EleutherAI/hendrycks_math)\n- **Split sizes**: Uses `train` or `test` split\n\n### Task\n- **Type**: single-turn\n- **Parser**: `ThinkParser` when `use_think=True` (default), else a basic `Parser` extracting the final boxed answer (`extract_boxed_answer`)\n- **Rubric overview**: Exact-match on parsed boxed answer (single criterion, weight 1.0)\n\n### Quickstart\nRun an evaluation with default settings:\n\n```bash\nuv run vf-eval hendrycksmath\n```\n\nConfigure model and sampling:\n\n```bash\nuv run vf-eval hendrycksmath \\\n  -m gpt-4.1-mini \\\n  -n 20 -r 3 -t 1024 -T 0.7\n```\n\nNotes:\n- Use `-a` / `--env-args` to pass environment-specific configuration as a JSON object.\n\n#### Use Training Split\n```bash\nuv run vf-eval hendrycksmath -a '{\"split\": \"train\"}'\n```\n\n#### Direct Reasoning Mode (No Thinking Tags)\n```bash\nuv run vf-eval hendrycksmath -a '{\"use_think\": false}'\n```\n\n#### Combined Configuration\n```bash\nuv run vf-eval hendrycksmath \\\n  -m gpt-4.1-mini \\\n  -n 50 -r 2 \\\n  -a '{\"split\": \"train\", \"use_think\": false}'\n```\n\n### Environment Arguments\n\n| Arg | Type | Default | Description |\n| --- | ---- | ------- | ----------- |\n| `use_think` | bool | `True` | Use thinking mode with `<think>` tags (`ThinkParser`) or direct reasoning (`Parser`) |\n| `system_prompt` | str | Auto | Custom system prompt. If None, uses `THINK_BOXED_SYSTEM_PROMPT` or `BOXED_SYSTEM_PROMPT` |\n| `split` | str | `\"test\"` | Dataset split to use: `\"train\"` (~7,500 problems) or `\"test\"` (~5,000 problems) |\n\n### Metrics\nSummarize key metrics your rubric emits and how they’re interpreted.\n\n| Metric | Meaning |\n| ------ | ------- |\n| `reward` | 1.0 if parsed boxed answer equals target, else 0.0 |","encoding":"utf-8","truncated":false,"total_bytes":2123},"status":null}