2026-09-05
The open dataset on GitHub is a snapshot, not a scrape dump
How to read the HumanoidPodium JSON/CSV on GitHub: same inclusion rule as the site, English field names, and why empty marks stay empty.
The live site and the GitHub snapshot share one rule: a public source_url is required for every result row. The repo is for people who want the table in JSON or CSV, not a second unofficial ranking invented by a scraper.
Field names stay in English on purpose so scripts do not break when the UI language changes. The website ships eight UI locales; the dataset does not silently translate marks or invent units.
If the site and a local clone disagree, trust the deployed pages after a refresh, then open an issue or use the contact page with a source URL. We will not merge unverified screenshots into the CSV.
Start from the rankings board if you are browsing, or from the GitHub README if you are wiring a notebook. Related Chinese product notes live on aiwenbiao.cn and are a different product surface.