Dynamic examples share IDs and URLs with dynamic-content tasks but have different acceptance contracts; the dynamic-content contracts are kept. State that the 530-task selection was documented after outcomes were available and is a rule-based cohort boundary, not a pre-outcome or representative selection. Label the 550-task export guidance as the historical v2.x edition. Archive/manifest rebuilt (source_revision a71b0b0f); make check: 37 tests OK, task_count 530.
Release v3.0.0: 530-task edition without duplicated examples
Make the 530-task population (definitions/populations/primary-content-20261103.json `cases`) the active edition v3.0.0: the earlier 430-task population plus 100 article tasks.
The 550-task catalog counted two 10-task example groups from the mixed webfetch suite:
- 10 `dynamic-examples` have the same IDs and URLs as tasks in the 100-task dynamic-content suite; counting them double-counts those targets. Their viewer config and JSONL are removed.
- 10 `challenge-examples` (Amazon, JD, Xiaohongshu, Zhihu, Douban, Dianping, Reddit, IMDb, Anthropologie, JustWatch list pages) are an example group. They stay as a separately reported challenge set, outside the counted edition.
Changes:
- manifest.json edition: tag v3.0.0, version 3.0.0, task_count 530, eight configurations, population_sha256 d6914e377dfc11c07627a733b3aa810a2b9fe7adf86e4acf85ad140680ecdd25 (VERSIONING.md method). Viewer metadata drops dynamic-examples and labels challenge-examples as uncounted.
- README/VERSIONING: v3.0.0 (530…
The GitHub repository lexmount/browser-eval is now lexmount/webstone. Prose and links only; manifest.json, snapshots.json and definitions/ are unchanged, so pinned revisions and hashes stay valid.
3 days ago
Release v2.2.0: corrected references for the 550-task edition (#7)
