Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -11,17 +11,23 @@ tags: [databases, query-equivalence, feature-engineering, rewrite-mining, SMT]
|
|
| 11 |
|
| 12 |
# VeBench
|
| 13 |
|
| 14 |
-
|
| 15 |
-
an MLIR-Style IR and SMT** (ICDE 2026)
|
| 16 |
-
|
|
|
|
|
|
|
| 17 |
|
| 18 |
-
| File | Component |
|
| 19 |
-
| --- | --- | --- |
|
| 20 |
-
| `
|
| 21 |
-
| `
|
| 22 |
-
| `
|
| 23 |
-
| `
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 24 |
|
| 25 |
-
|
| 26 |
-
|
| 27 |
-
`REPRODUCE.md` for provenance, licenses, and regeneration.
|
|
|
|
| 11 |
|
| 12 |
# VeBench
|
| 13 |
|
| 14 |
+
The labeled benchmark for **VeRA: Mining and Verifying Feature-Pipeline Rewrites
|
| 15 |
+
with an MLIR-Style IR and SMT** (ICDE 2026): nine components for verified
|
| 16 |
+
feature-pipeline rewrite mining. Generated components regenerate from fixed seeds
|
| 17 |
+
via the artifact (`scripts/export_bench.py`); real-code components ship as
|
| 18 |
+
license-stamped samples.
|
| 19 |
|
| 20 |
+
| File | Component | Size | Origin |
|
| 21 |
+
| --- | --- | --- | --- |
|
| 22 |
+
| `sql_mining_corpus.jsonl` | SQL mining corpus | 300 | generated |
|
| 23 |
+
| `dataframe_mining_corpus.jsonl` | dataframe mining corpus | 300 | generated |
|
| 24 |
+
| `soundness_rules.jsonl` | soundness rule set | 31 | hand-built |
|
| 25 |
+
| `mutation_rules.jsonl` | mutation rule set | 400 | mechanical |
|
| 26 |
+
| `hf_jupyter_sample.jsonl` | real-notebook scan | 1500 | public GitHub |
|
| 27 |
+
| `feast_sample.jsonl` | Feast feature definitions | 20 | Feast templates |
|
| 28 |
+
| `job_sample.jsonl` | JOB coverage suite | 113 | IMDb (JOB) |
|
| 29 |
+
| `tpch_coverage.jsonl` | TPC-H coverage suite | 22 | TPC-H |
|
| 30 |
+
| `tpcds_coverage.jsonl` | TPC-DS coverage suite | 99 | TPC-DS |
|
| 31 |
|
| 32 |
+
See the artifact `README.md` and `REPRODUCE.md` for provenance, licenses, labels
|
| 33 |
+
(lowering status, admission tier, guard, verdict), and regeneration.
|
|
|