```json { "table_name": "neal_reading_groups", "table_type_tags": [ {"tag": "dimension", "confidence": 0.75}, {"tag": "misc", "confidence": 0.25} ], "primary_type": "dimension", "insufficient_evidence": false, "entity": { "columns": ["reading_id", "rank"], "unique_per_row": true }, "time_axis": null, "reasoning": "The composite entity (reading_id, rank) is unique per row — reading_id Uniqueness 0.273 with RPUV 3.67, rank Uniqueness 0.018 with RPUV 55.00, and their product (60 × 4 = 240) matches the row count. Each row describes a ranked group breakdown for a reading (label + value pairs, with remainder_count on rank=4 rows). No time axis exists — the provenance columns are synthetic. The rows are descriptive attributes of reading-rank combinations, not actions or state snapshots → dimension. The opaque schema (reading_id references an external context, label values vary wildly) adds minor misc uncertainty." } ```