```json { "classifications": [ {"column": "id_runs", "role": "identifier", "confidence": 1.0, "reason": "Primary key with unique alphanumeric codes for each run"}, {"column": "task", "role": "dimension", "confidence": 1.0, "reason": "Categorical task name used for grouping and filtering (14 distinct values)"}, {"column": "model", "role": "dimension", "confidence": 1.0, "reason": "Model name used for grouping and filtering (311 distinct values)"}, {"column": "Best score (across scorers)", "role": "measure", "confidence": 1.0, "reason": "Numeric score metric to be aggregated (0.0 to 1.0 range)"}, {"column": "Scores", "role": "attribute", "confidence": 0.95, "reason": "Formatted score strings with scorer names, displayed but not aggregated"}, {"column": "log viewer", "role": "attribute", "confidence": 1.0, "reason": "URL for display purposes, not grouped or aggregated"}, {"column": "logs", "role": "attribute", "confidence": 1.0, "reason": "S3 URL for display purposes, not grouped or aggregated"}, {"column": "job", "role": "identifier", "confidence": 0.95, "reason": "Job identifier codes used for tracking (914 distinct values)"}, {"column": "started_at", "role": "dimension", "confidence": 1.0, "reason": "Timestamp for filtering and time-based grouping"}, {"column": "Status", "role": "dimension", "confidence": 1.0, "reason": "Categorical status field for filtering (only 'Success' present)"}, {"column": "Manual evaluation", "role": "dimension", "confidence": 0.9, "reason": "Boolean flag for filtering (99.5% null, single value 'true')"}, {"column": "task version", "role": "dimension", "confidence": 1.0, "reason": "Version codes for grouping and filtering (32 distinct values)"}, {"column": "billable_input_tokens", "role": "measure", "confidence": 0.7, "reason": "Token count field (currently empty but intended as measure)"}, {"column": "billable_output_tokens", "role": "measure", "confidence": 0.7, "reason": "Token count field (currently empty but intended as measure)"}, {"column": "cache_read_tokens", "role": "measure", "confidence": 0.7, "reason": "Token count field (currently empty but intended as measure)"}, {"column": "cache_write_tokens", "role": "measure", "confidence": 0.7, "reason": "Token count field (currently empty but intended as measure)"}, {"column": "comments", "role": "attribute", "confidence": 1.0, "reason": "Free-form descriptive text for display"}, {"column": "id_model_version", "role": "identifier", "confidence": 1.0, "reason": "Foreign key linking to model version (311 distinct values)"}, {"column": "Model name", "role": "dimension", "confidence": 1.0, "reason": "Model name for grouping (167 distinct values)"}, {"column": "Version release date", "role": "dimension", "confidence": 1.0, "reason": "Date field for temporal grouping and filtering"}, {"column": "Hugging Face developer id", "role": "dimension", "confidence": 1.0, "reason": "Developer identifier for grouping (14 distinct values)"}, {"column": "Link", "role": "attribute", "confidence": 1.0, "reason": "URL for display purposes"}, {"column": "Notes", "role": "attribute", "confidence": 1.0, "reason": "Free-form descriptive text for display"}, {"column": "benchmarks/runs", "role": "attribute", "confidence": 0.9, "reason": "Comma-separated list of run IDs for display"}, {"column": "Display name", "role": "dimension", "confidence": 1.0, "reason": "Display label for grouping (185 distinct values)"}, {"column": "Unique display name", "role": "dimension", "confidence": 1.0, "reason": "Unique display label for grouping (194 distinct values)"}, {"column": "Domain", "role": "dimension", "confidence": 1.0, "reason": "Categorical domain field for grouping (21 distinct values)"}, {"column": "Model task", "role": "dimension", "confidence": 1.0, "reason": "Task categories for grouping (74 distinct values)"}, {"column": "Organization", "role": "dimension", "confidence": 1.0, "reason": "Organization name for grouping (21 distinct values)"}, {"column": "Authors", "role": "attribute", "confidence": 1.0, "reason": "Long text list of authors for display"}, {"column": "Publication date", "role": "dimension", "confidence": 1.0, "reason": "Date field for temporal grouping and filtering"}, {"column": "Reference", "role": "attribute", "confidence": 1.0, "reason": "Descriptive reference text for display"}, {"column": "Link_model", "role": "attribute", "confidence": 1.0, "reason": "URL for display purposes"}, {"column": "Citations", "role": "measure", "confidence": 1.0, "reason": "Numeric citation count to be aggregated"}, {"column": "Notability criteria", "role": "dimension", "confidence": 1.0, "reason": "Categorical criteria for grouping (16 distinct values)"}, {"column": "Notability criteria notes", "role": "attribute", "confidence": 1.0, "reason": "Free-form descriptive text for display"}, {"column": "Parameters", "role": "measure", "confidence": 1.0, "reason": "Numeric parameter count to be aggregated"}, {"column": "Parameters notes", "role": "attribute", "confidence": 1.0, "reason": "Free-form descriptive text for display"}, {"column": "Training compute (FLOP)", "role": "measure", "confidence": 1.0, "reason": "Numeric FLOP count to be aggregated"}, {"column": "Training compute notes", "role": "attribute", "confidence": 1.0, "reason": "Free-form descriptive text for display"}, {"column": "Training dataset", "role": "dimension", "confidence": 1.0, "reason": "Dataset name for grouping (7 distinct values)"}, {"column": "Training dataset notes", "role": "attribute", "confidence": 1.0, "reason": "Free-form descriptive text for display"}, {"column": "Training dataset size (datapoints)", "role": "measure", "confidence": 1.0, "reason": "Numeric datapoint count to be aggregated"}, {"column": "Dataset size notes", "role": "attribute", "confidence": 1.0, "reason": "Free-form descriptive text for display"}, {"column": "Batch size", "role": "measure", "confidence": 1.0, "reason": "Numeric batch size to be aggregated"}, {"column": "Batch size notes", "role": "attribute", "confidence": 1.0, "reason": "Free-form descriptive text for display"}, {"column": "Epochs", "role": "measure", "confidence": 1.0, "reason": "Numeric epoch count to be aggregated"}, {"column": "Training hardware", "role": "dimension", "confidence": 1.0, "reason": "Hardware type for grouping (11 distinct values)"}, {"column": "Hardware quantity", "role": "measure", "confidence": 1.0, "reason": "Numeric hardware count to be aggregated"}, {"column": "Abstract", "role": "attribute", "confidence": 1.0, "reason": "Long descriptive text for display"}, {"column": "Model accessibility", "role": "dimension", "confidence": 1.0, "reason": "Categorical accessibility type for grouping (7 distinct values)"}, {"column": "Hardware utilization", "role": "measure", "confidence": 1.0, "reason": "Numeric utilization rate to be aggregated"}, {"column": "Confidence", "role": "dimension", "confidence": 1.0, "reason": "Categorical confidence level for grouping (5 distinct values)"}, {"column": "Base model", "role": "dimension", "confidence": 1.0, "reason": "Base model name for grouping (18 distinct values)"}, {"column": "Finetune compute (FLOP)", "role": "measure", "confidence": 1.0, "reason": "Numeric FLOP count to be aggregated"}, {"column": "Finetune compute notes", "role": "attribute", "confidence": 1.0, "reason": "Free-form descriptive text for display"}, {"column": "Training code accessibility", "role": "dimension", "confidence": 1.0, "reason": "Categorical accessibility type for grouping (4 distinct values)"}, {"column": "Accessibility notes", "role": "attribute", "confidence": 1.0, "reason": "Free-form descriptive text for display"}, {"column": "Training compute cost (2023 USD)", "role": "measure", "confidence": 1.0, "reason": "Numeric cost amount to be aggregated"}, {"column": "Training compute estimation method", "role": "dimension", "confidence": 1.0, "reason": "Categorical method for grouping (13 distinct values)"}, {"column": "Country (of organization)", "role": "dimension", "confidence": 1.0, "reason": "Country name for grouping (4 distinct values)"}, {"column": "mean_score", "role": "measure", "confidence": 1.0, "reason": "Numeric mean score to be aggregated"}, {"column": "stderr", "role": "measure", "confidence": 1.0, "reason": "Numeric standard error to be aggregated"}, {"column": "best_score", "role": "measure", "confidence": 1.0, "reason": "Numeric best score to be aggregated"}, {"column": "original_task_name", "role": "dimension", "confidence": 1.0, "reason": "Original task name for grouping (16 distinct values)"} ] } ```