Revise article scoring and publish HTML leaderboards

[bingran-you]

Summary

• replace the arithmetic-mean primary rank with IQM while retaining mean and median diagnostics
• publish native raw-score task leaderboards and a positive IQM + 100 display index for the final plot
• render all website leaderboards as responsive HTML/CSS from synchronized JSON rather than raster images
• document exact model routes and provider-specific reasoning settings (max / xhigh)
• preserve the latest main hero artwork and nine-task navigator

Inference settings

Model	Route	Reasoning setting
GPT-5.6 Sol	azure/gpt-5.6-sol	max
GPT-5.5	azure/gpt-5.5	xhigh
Claude Opus 4.8	anthropic/claude-opus-4-8 via Claude OAuth	max
GPT-5.4 Mini	azure/gpt-5.4-mini	xhigh

These are provider-specific categorical labels, not a shared numeric compute scale.

Validation

• uv run pytest -q — 92 passed, 18 skipped
• uv run ruff check .
• uv build
• uv run python scripts/validate_tasks.py — all nine tasks valid
• desktop and mobile browser checks: 9 task charts, 4 final rows, no chart images, no overflow or console errors

[vercel]

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

Project	Deployment	Actions	Updated (UTC)
genesis-bench	Ready	Preview, Comment	Jul 13, 2026 10:28pm

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: e032bfc483

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Handle final display scores below zero

For any future model whose IQM is <= -100, which is allowed because the normalized scores are explicitly unbounded, this + 100 transform produces a non-positive positive_display_score. The new HTML and PNG final leaderboards both treat this field as a bar width/value from a zero lower bound, so a valid very-low-scoring model would be clipped or omitted instead of showing its meaningful negative gap; use a plotting domain that can include negative display values or derive an offset that is guaranteed for the current cohort.

Useful? React with 👍 / 👎.