Update README.md

Revert README to previous wording (undo 14,419 edit)

Snow: align schema space with the deployed inference graphs (max 14,419 grouped columns, ACS_2); gold re-expressed in that space, n_schema_columns == len(schema), 0 unresolved / 0 empty

README: note sqlite_sequence exclusion in Spider schema lists (verified vs official tables.json)

One consistent grouped-node space for Spider 2.0: schemas as flat node lists, n_schema_columns == len(schema), gold re-expressed in system space (31 snow CRYPTO remaps, lite family space + SQL-disambiguated shards); strict exact-membership checker

Add CSV twins of all parquet comparison files for offline viewing

Case-by-case: add gold_sql + both gold sets + explicit extra/missing diff columns for each side

README: simplify PK-convention step wording

Comparison folder -> comparison_with_prior_golds, parquet format (HF blob page renders parquet as interactive table)

README: gold-construction as concise pipeline arrows

Slim case-by-case CSVs under 1MB for HF blob preview; official sample ids (question_id/instance_id) + n_schema_columns in eval sets; unified o4-mini gold-construction description; document schemas/ join

Fix lite-256 gold for 94 dual-gold instances (use adjudicated lite-SQL-derived key) + join-key completeness patch (local003, local039, sf014); verified against prior released lite set

Add Spider 2.0-Lite 256 + gold checker + schemas; rename comparison to prior-golds-by-lexical-matches; restructure README

Simplify: evaluation sets + KaSLA case-by-case only; short README linking to github.com/thanhdath/grast-sql

Add Spider/BIRD/Spider2.0-Snow(256) schema-linking eval sets + KaSLA case-by-case comparison

initial commit