- Added empty string docstring column as placeholder
- Ensures compatibility when training across all Coq/Lean4 datasets
- Real docstrings to be extracted and added in future update
- Prevents training errors from schema mismatch
This dataset had 2.2% docstring coverage, which is below the 5%
minimum threshold for retaining the column. Sparse docstring columns
(<5% populated) provide insufficient signal for ML training while adding
schema complexity and storage overhead. The column has been removed to
maintain data quality standards across the theorem prover dataset collection.
Initial upload: 52,929 declarations from Verified Software Toolchain
