Skip to content

Latest commit

 

History

History
104 lines (81 loc) · 3.69 KB

File metadata and controls

104 lines (81 loc) · 3.69 KB

Changelog

All notable changes to VulnGym are documented here. The format follows Keep a Changelog and this project adheres to Semantic Versioning.

[0.1.4] — 2026-06-26

Data refresh — further expansion of human-audited coverage.

Changed

  • Human-audited entries grew from 350 → 393 / 408 (96.3 %), covering 178 / 184 advisories (96.7 %).
  • Updated human-audit status flags only; row counts, schema, desc coverage, and vulnerability-type distribution are unchanged.

Stats

  • reports: 184 (unchanged)
  • entries: 408 (unchanged)
  • human-audited entries (verify = 1): 393 (was 350)
  • human-audited advisories (≥ 1 verified entry): 178 (was 163)

[0.1.3] — 2026-06-18

Data refresh — further expansion of human-audited coverage and broad annotation refinements.

Changed

  • Human-audited entries grew from 274 → 350 / 408 (85.8 %), covering 163 / 184 advisories (88.6 %).

Added

  • desc field on the entry_point, critical_operation, and trace nodes of 400 entries — a natural-language explanation of each node's role in the vulnerability chain.

Stats

  • reports: 184 (unchanged)
  • entries: 408 (unchanged)
  • human-audited entries (verify = 1): 350 (was 274)
  • human-audited advisories (≥ 1 verified entry): 163 (was 137)

[0.1.2] — 2026-05-31

Data refresh — significant expansion of human-audited coverage and annotation refinements.

Changed

  • Human-audited entries grew from 113 → 274 / 408 (67.2 %), covering 137 / 184 advisories (74.5 %).
  • Refined entry_point, critical_operation, and trace annotations on 80 entries for improved accuracy.

Stats

  • reports: 184 (unchanged)
  • entries: 408 (unchanged)
  • human-audited entries (verify = 1): 274 (was 113)
  • human-audited advisories (≥ 1 verified entry): 137 (was 61)

[0.1.1] — 2026-05-15

Data refresh — adds a human-audit flag and additional human-verified entries.

Added

  • verify field on every row in data/entries.jsonl (int, 0 or 1): 1 marks entries that have been reviewed and confirmed by a human annotator (high-confidence ground truth); 0 marks automatically annotated entries that have not yet been human-confirmed.
  • 113 / 408 entries (≈ 27.7 %) are now flagged verify = 1, covering 61 / 184 advisories (≈ 33.2 %; 50 advisories have all of their entries verified, 11 are partially verified).

Changed

  • Refined values of selected entry_point, critical_operation, trace, and other annotation fields in data/entries.jsonl. Row counts and the report_identry_id join structure are unchanged.
  • SCHEMA.md now documents verify; human_confirmed is removed from the "intentionally omitted internal fields" invariant since the audit status is exposed publicly via verify.

Stats

  • reports: 184 (unchanged)
  • entries: 408 (unchanged)
  • human-audited entries (verify = 1): 113
  • human-audited advisories (≥ 1 verified entry): 61

[0.1.0] — 2026-05-07

Initial open-source release.

Added

  • data/reports.jsonl — 184 GitHub Advisories (report-level aggregates).
  • data/entries.jsonl — 408 per-entry-point records with entry_point / critical_operation / trace annotations.
  • SCHEMA.md — full field reference and invariants.
  • examples/load_dataset.py — stdlib / pandas / HuggingFace datasets loaders.
  • examples/evaluate.py — coverage / recall evaluator.
  • examples/example_result.jsonl — illustrative tool-findings submission.
  • CC-BY-4.0 license.

Stats

  • reports: 184
  • entries: 408
  • projects: 38
  • repositories: 23