Remove abandoned training-time caps

Keep only the paper dataset release

Publish unbounded v1 CPDP datasets and splits

Add all v1 CPDP split files

Add v1 Defactors CPDP split (1700 records)

Remove obsolete Devign README reference

Update README for four-dataset releases

Clarify sample limits without cap terminology

Remove cap documentation

Remove cap artifacts from main

Regenerate v1 training caps for uncapped ASTs

Regenerate v1 Defactors without AST node cap

Regenerate v1 BugsInPy without AST node cap

Document v2.0.0 and v2.0.1 policies

Release v2.0.1 datasets and regenerated caps

Document v2 source-normalization retention rules

Release v2.0.0 comment-free datasets and regenerated caps

Rebuild v2 with direct comment removal

Update BugsInPy cap summary

Document invalid BugsInPy bag filtering

Remove invalid BugsInPy placeholder bags

Replace v2 datasets with comment-free reconstruction

Keep v2 limited to four reconstructed MIL datasets

Publish v2.0.0 reconstructed datasets and matching caps

Add source replay audit reports

Document original dataset sources

Add dataset card metadata

Add dataset statistics and construction audit

Rebuild Promise and Defactors with complete bags

sync latest cleaned AST dataset: bigvul_c.jsonl

sync latest cleaned AST dataset: bugsinpy_python.jsonl

sync latest cleaned AST dataset: defactors_python.jsonl

sync latest cleaned AST dataset: devign_c.jsonl

sync latest cleaned AST dataset: promise_java.jsonl

reupload unified AST dataset: bigvul_c.jsonl

reupload unified AST dataset: bugsinpy_python.jsonl

reupload unified AST dataset: defactors_python.jsonl

reupload unified AST dataset: devign_c.jsonl

reupload unified AST dataset: promise_java.jsonl

update indexed datasets

update indexed datasets

Add hamnet datasets via LFS

initial commit