Spaces:
Sleeping
Sleeping
| name: data-cleaning-openenv | |
| version: 1.0.0 | |
| description: > | |
| An OpenEnv-compliant environment where AI agents learn to clean | |
| messy real-world datasets step by step. Tasks include duplicate | |
| removal, missing value imputation, dtype fixing, outlier detection, | |
| and schema validation. | |
| author: thorodin103 | |
| license: mit | |
| tags: | |
| - openenv | |
| - data-cleaning | |
| - real-world | |
| - pandas | |
| - tabular | |
| - reinforcement-learning | |
| - agent | |
| api: | |
| base_url: https://thorodin103-data-cleaning-openenv.hf.space | |
| reset: POST /reset | |
| step: POST /step | |
| state: GET /state | |
| reset_task_scoped: POST /reset/{task_id} | |
| step_task_scoped: POST /step/{task_id} | |
| state_task_scoped: GET /state/{task_id} | |
| tasks: GET /tasks | |
| validate: GET /validate | |
| health: GET /health | |
| tasks: | |
| - id: easy_dedup_rename | |
| difficulty: easy | |
| description: Remove duplicate rows and rename columns to snake_case | |
| max_steps: 10 | |
| scoring: | |
| duplicate_score: 0.5 | |
| schema_score: 0.5 | |
| - id: medium_missing_dtype | |
| difficulty: medium | |
| description: Fill missing values and fix wrong data types | |
| max_steps: 15 | |
| scoring: | |
| missing_score: 0.5 | |
| dtype_score: 0.5 | |
| - id: hard_full_pipeline | |
| difficulty: hard | |
| description: Full pipeline — duplicates, missing, dtypes, outliers, schema | |
| max_steps: 20 | |
| scoring: | |
| duplicate_score: 0.2 | |
| missing_score: 0.2 | |
| dtype_score: 0.2 | |
| outlier_score: 0.2 | |
| schema_score: 0.2 | |
| observation_space: | |
| task_id: string | |
| step: integer | |
| dataset_info: object | |
| columns: list | |
| shape: list | |
| missing_values: object | |
| dtypes: object | |
| duplicate_count: integer | |
| sample_rows: list | |
| available_operations: list | |
| task_description: string | |
| message: string | |
| action_space: | |
| operation: | |
| type: string | |
| values: | |
| - remove_duplicates | |
| - fill_missing | |
| - fix_dtype | |
| - remove_outliers | |
| - rename_columns | |
| - validate_schema | |
| - finish | |
| parameters: | |
| type: object | |
| reward_space: | |
| total: | |
| type: float | |
| range: [0.0, 1.0] | |
| components: | |
| - duplicate_score | |
| - missing_score | |
| - dtype_score | |
| - outlier_score | |
| - schema_score | |
| - penalty | |
| baseline_scores: | |
| model: gpt-4o-mini | |
| description: Sequence-aware reward function (rewards optimal order of operations) | |
| easy_dedup_rename: 0.9900 | |
| medium_missing_dtype: 0.7000 | |
| hard_full_pipeline: 0.6636 | |
| average: 0.7845 | |
| runtime: | |
| language: python | |
| version: "3.10" | |
| framework: fastapi | |
| port: 7860 | |