```json { "classifications": [ {"column": "id_runs", "label": "none", "confidence": 1.0, "reason": "Unique identifier for benchmark runs"}, {"column": "task", "label": "none", "confidence": 1.0, "reason": "Task name/type, categorical dimension"}, {"column": "model", "label": "none", "confidence": 1.0, "reason": "Model name, categorical dimension"}, {"column": "Best score (across scorers)", "label": "none", "confidence": 1.0, "reason": "Numeric metric/measure"}, {"column": "Scores", "label": "none", "confidence": 1.0, "reason": "Score values with uncertainty, metric"}, {"column": "log viewer", "label": "none", "confidence": 1.0, "reason": "URL reference"}, {"column": "logs", "label": "none", "confidence": 1.0, "reason": "URL reference"}, {"column": "job", "label": "none", "confidence": 1.0, "reason": "Job identifier"}, {"column": "started_at", "label": "temporal", "confidence": 1.0, "reason": "Timestamp of when benchmark run started"}, {"column": "Status", "label": "none", "confidence": 1.0, "reason": "Status indicator, categorical"}, {"column": "Manual evaluation", "label": "none", "confidence": 1.0, "reason": "Boolean flag"}, {"column": "task version", "label": "none", "confidence": 1.0, "reason": "Version identifier"}, {"column": "billable_input_tokens", "label": "none", "confidence": 1.0, "reason": "Numeric measure (empty in data)"}, {"column": "billable_output_tokens", "label": "none", "confidence": 1.0, "reason": "Numeric measure (empty in data)"}, {"column": "cache_read_tokens", "label": "none", "confidence": 1.0, "reason": "Numeric measure (empty in data)"}, {"column": "cache_write_tokens", "label": "none", "confidence": 1.0, "reason": "Numeric measure (empty in data)"}, {"column": "comments", "label": "none", "confidence": 1.0, "reason": "Text notes/comments"}, {"column": "id_model_version", "label": "none", "confidence": 1.0, "reason": "Model version identifier"}, {"column": "Model", "label": "none", "confidence": 1.0, "reason": "Model name (empty in data)"}, {"column": "Version release date", "label": "temporal", "confidence": 1.0, "reason": "Date when model version was released"}, {"column": "Hugging Face developer id", "label": "none", "confidence": 1.0, "reason": "Developer identifier"}, {"column": "Link", "label": "none", "confidence": 1.0, "reason": "URL reference"}, {"column": "Notes", "label": "none", "confidence": 1.0, "reason": "Text notes/descriptions"}, {"column": "benchmarks/runs", "label": "none", "confidence": 1.0, "reason": "List of run identifiers"}, {"column": "Display name", "label": "none", "confidence": 1.0, "reason": "Display name for model"}, {"column": "Unique display name", "label": "none", "confidence": 1.0, "reason": "Unique display name for model"}, {"column": "Domain", "label": "none", "confidence": 1.0, "reason": "Domain categories, categorical dimension"}, {"column": "Task", "label": "none", "confidence": 1.0, "reason": "Task description (empty in data)"}, {"column": "Organization", "label": "none", "confidence": 1.0, "reason": "Organization name, categorical dimension"}, {"column": "Authors", "label": "none", "confidence": 1.0, "reason": "Author names, text attribute"}, {"column": "Publication date", "label": "temporal", "confidence": 1.0, "reason": "Date when model was published"}, {"column": "Reference", "label": "none", "confidence": 1.0, "reason": "Reference text/description"}, {"column": "Link_model", "label": "none", "confidence": 1.0, "reason": "URL reference"}, {"column": "Citations", "label": "none", "confidence": 1.0, "reason": "Citation count, numeric measure"}, {"column": "Notability criteria", "label": "none", "confidence": 1.0, "reason": "Criteria categories"}, {"column": "Notability criteria notes", "label": "none", "confidence": 1.0, "reason": "Text notes"}, {"column": "Parameters", "label": "none", "confidence": 1.0, "reason": "Model parameter count, numeric measure"}, {"column": "Parameters notes", "label": "none", "confidence": 1.0, "reason": "Text notes"}, {"column": "Training compute (FLOP)", "label": "none", "confidence": 1.0, "reason": "Compute measure in FLOP"}, {"column": "Training compute notes", "label": "none", "confidence": 1.0, "reason": "Text notes"}, {"column": "Training dataset", "label": "none", "confidence": 1.0, "reason": "Dataset name, categorical"}, {"column": "Training dataset notes", "label": "none", "confidence": 1.0, "reason": "Text notes"}, {"column": "Training dataset size (datapoints)", "label": "none", "confidence": 1.0, "reason": "Dataset size, numeric measure"}, {"column": "Dataset size notes", "label": "none", "confidence": 1.0, "reason": "Text notes"}, {"column": "Batch size", "label": "none", "confidence": 1.0, "reason": "Batch size, numeric measure"}, {"column": "Batch size notes", "label": "none", "confidence": 1.0, "reason": "Text notes"}, {"column": "Epochs", "label": "none", "confidence": 1.0, "reason": "Number of epochs, numeric measure"}, {"column": "Training hardware", "label": "none", "confidence": 1.0, "reason": "Hardware type, categorical"}, {"column": "Hardware quantity", "label": "none", "confidence": 1.0, "reason": "Hardware count, numeric measure"}, {"column": "Abstract", "label": "none", "confidence": 1.0, "reason": "Text abstract/description"}, {"column": "Model accessibility", "label": "none", "confidence": 1.0, "reason": "Accessibility type, categorical"}, {"column": "Hardware utilization", "label": "none", "confidence": 1.0, "reason": "Utilization rate, numeric measure"}, {"column": "Confidence", "label": "none", "confidence": 1.0, "reason": "Confidence level, categorical"}, {"column": "Base model", "label": "none", "confidence": 1.0, "reason": "Base model name, categorical"}, {"column": "Finetune compute (FLOP)", "label": "none", "confidence": 1.0, "reason": "Compute measure in FLOP"}, {"column": "Finetune compute notes", "label": "none", "confidence": 1.0, "reason": "Text notes"}, {"column": "Training code accessibility", "label": "none", "confidence": 1.0, "reason": "Accessibility type, categorical"}, {"column": "Accessibility notes", "label": "none", "confidence": 1.0, "reason": "Text notes"}, {"column": "Training compute cost (2023 USD)", "label": "none", "confidence": 1.0, "reason": "Cost measure in USD"}, {"column": "Training compute estimation method", "label": "none", "confidence": 1.0, "reason": "Method description, categorical"}, {"column": "Country (of organization)", "label": "spatial", "confidence": 0.9, "reason": "Country names, geographic location but not used for plotting/joining"}, {"column": "mean_score", "label": "none", "confidence": 1.0, "reason": "Mean score, numeric measure"}, {"column": "stderr", "label": "none", "confidence": 1.0, "reason": "Standard error, numeric measure"}, {"column": "best_score", "label": "none", "confidence": 1.0, "reason": "Best score, numeric measure"}, {"column": "original_task_name", "label": "none", "confidence": 1.0, "reason": "Task name, categorical dimension"} ] } ```