{"data":{"kind":"file","path":"README.md","version_id":"ptcbdpqs2bny6xwjuizj3koz","entry":{"name":"README.md","path":"README.md","is_directory":false,"size":1465,"modified_at":"2025-08-28T22:20:20.284000","content_hash":"1b368cd708a9a3af27fd220b597dea96b7125931344462249dd6b716be622fbe"},"entries":[],"content":"# chess\n\nThis environment functions as both an environment and a benchmark for LLM's to play chess against open source stockfish.\nCurrently, version 0.1, the agent will play as white.\n\n### Overview\n- **Environment ID**: `vf_chess_eval`\n- **Short description**: Simulates a chess game versus Stockfish\n- **Tags**: eval, chess\n\n### Datasets\n- **Primary dataset(s)**: No dataset, starting from empty positions\n\n### Task\n- **Type**: multi-turn\n- **Parser**: custom\n- **Rubric overview**: stockfish_reward. Using stockfish's evaluation, we compare each move of the agent with the optimal move in the position.\nThe reward is proportional to the change in evaluation, with a cutoff.\n\n### Quickstart\nRun an evaluation with default settings:\n\n```bash\nuv run vf-eval chess\n```\n\nConfigure model and sampling:\n\n```bash\nuv run vf-eval chess   -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### Environment Arguments\nDocument any supported environment arguments and their meaning. Example:\n\n| Arg | Type | Default | Description |\n| --- | ---- | ------- | ----------- |\n| `engine_path` | str | `\"stockfish\"` | Path to stockfish executable  |\n| `stockfish_depth` | int | `12` | Limit on how deep stockfish goes into lines |\n\n### Metrics\n\n| Metric | Meaning |\n| ------ | ------- |\n| `reward` | Rewards, the closer to 0 the worse |\n\n","encoding":"utf-8","truncated":false,"total_bytes":1465},"status":null}