wallfacers/weft-script-lineage-synth
weft-script-lineage — synthetic training data + negative-result evidence Companion data for the model wallfacers/weft-lineage-extractor-1.5b, a research artifact demonstrating that synthetic-only training induces a memorization leak in small models for ETL table-lineage extraction. Contents train.jsonl / heldout.jsonl — 10,000 + 600 synthetic ETL scripts with table-lineage labels (Python/Shell). out-jvm/ adds the Scala/Java-augmented variant. reports/ — the… See the full description on the dataset page: https://huggingface.co/datasets/wallfacers/weft-script-lineage-synth.
Upload reports/paper-negative-result-findings.md with huggingface_hub
Upload reports/ADJUDICATION.md with huggingface_hub
Upload reports/leak-curve.md with huggingface_hub
Upload reports/leak-report-jvm.md with huggingface_hub
Upload reports/leak-report.md with huggingface_hub
Upload reports/eval-real-jvm.md with huggingface_hub
Upload reports/eval-real.md with huggingface_hub
Upload reports/eval-report-v2.md with huggingface_hub
041-R v2: JVM-augmented synthetic
Upload README.md with huggingface_hub
Upload eval-report.md with huggingface_hub
041 v2: train+heldout synth data
initial commit
