{"data":{"kind":"file","path":"README.md","version_id":"dmaiit68emz0jyhia5o7isr9","entry":{"name":"README.md","path":"README.md","is_directory":false,"size":462,"modified_at":"2026-04-06T23:58:08.965000","content_hash":"1016f8bb40693ffe2953ce6a20a31dff8966c1d4f518628f5db79d127eb389b0"},"entries":[],"content":"# hanabi-simple\n\nHanabi training environment for Prime Intellect Hosted Training.\n\nThis environment loads Hanabi JSONL data from a local file or a Hugging Face\ndataset repo and scores model outputs with two rewards:\n\n- state tracking (`deduction`)\n- move quality (`action` against `move_ratings`)\n\nExpected dataset fields:\n\n- `user_prompt_training`\n- `move_ratings`\n- `programmatic_deduction`\n\nThe environment entrypoint is `load_environment(dataset_path=...)`.\n","encoding":"utf-8","truncated":false,"total_bytes":462},"status":null}