Create README.md

README.md

@@ -0,0 +1,10 @@
+---
+license: mit
+task_categories:
+- table-question-answering
+pretty_name: FreeformTableQA Selection
+size_categories:
+- n<1K
+---
+
+This dataset represents a curated subset of the FreeformTableQA collection, refined to provide a more rigorous benchmark for table-based reasoning. While the original dataset covers a broad range of "free-form" queries—which often include complex, semi-structured, or non-grid layouts—it also contains instances of noise and misaligned labels. To ensure higher evaluation accuracy, this version features 100 manually verified examples where the natural language queries and tabular evidence have been audited for logical consistency and factual correctness. This selection is intended for high-precision testing of Large Language Models (LLMs) in interpreting unstructured tabular data.