Publish per-task article leaderboards

[bingran-you]

Summary

• keep the top-level README focused on the final nine-task average leaderboard
• replace the offline wide table with nine independently ranked task leaderboards followed by the final average leaderboard
• add score-detail JSON links to every task leaderboard row
• encode the same ordered 10-leaderboard structure in leaderboard/article_suite.json
• refresh score-compatible task-digest provenance without changing any published scores

Validation

• uv run pytest -q — 91 passed, 18 skipped
• uv run ruff check .
• uv build
• uv run python scripts/validate_tasks.py — all nine tasks valid
• leaderboard regeneration is deterministic
• published per-task and average scores are unchanged

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
genesis-bench	Ignored		Jul 13, 2026 5:46pm