Skip to content

CONTRIBUTING.md

Latest commit

 

History

History
63 lines (46 loc) · 2.48 KB

CONTRIBUTING.md

File metadata and controls

63 lines (46 loc) · 2.48 KB
Raw

Contributing

Thanks for helping maintain Awesome Chinese AI Models.

What belongs here

Accepted:

  • Chinese AI models or model families.
  • Models with strong Chinese-language or China ecosystem relevance.
  • Official model docs, APIs, model cards, papers, changelogs, and pricing pages.
  • Open-source weights, inference recipes, quantization releases, and deployment guides.
  • Benchmarks with clear source and test conditions.
  • Reproducible playbooks, tutorials, demos, and workflows.
  • Major Chinese AI ecosystem news: releases, API changes, funding, partnerships, policy, safety, infrastructure.

Not accepted by default:

  • Unsourced rumors.
  • Pure marketing pages without technical detail.
  • Random model mirrors.
  • SEO farms or AI-generated repost sites.
  • Gray-market automation, account abuse, cracking, scraping abuse, copyright evasion, or deepfake fraud.
  • Benchmark claims without source, version, and test condition.

Source priority

P0 — official / primary:

  • Official website, blog, docs, changelog, pricing page.
  • Official GitHub, Hugging Face, ModelScope, OpenXLab, Paddle AI Studio.
  • Technical report, arXiv paper, whitepaper, model card.

P1 — trusted third-party:

  • OpenCompass, C-Eval, SuperCLUE, CMMLU, MTEB/C-MTEB, BFCL, benchmark reports.
  • Credible technical media or research organizations.
  • Reproducible GitHub projects and demos.

P2 — community signal:

  • X/Twitter, 即刻, 知乎, Bilibili, 小红书, Discord/Slack/飞书群讨论.
  • These are discovery signals only. Please find a P0/P1 source before adding stable entries.

PR checklist

  • The entry has a source URL.
  • The category is clear.
  • The description is factual and neutral.
  • New model entries include provider, model type, access mode, and license if open weight.
  • Benchmark entries include model version, benchmark name, source, and test conditions.
  • Commercial/self-submitted entries disclose affiliation.
  • No generated README block was edited manually unless generated by script.
  • python3 scripts/validate_data.py passes.
  • python3 scripts/generate_readme.py --check passes.

Style

  • Prefer Chinese descriptions with English model names preserved.
  • Keep bullets short: one sentence is ideal.
  • Avoid hype words: 最强, 吊打, 遥遥领先, 完全免费无限用.
  • Attribute claims: “official claim”, “third-party benchmark”, “community reproduction”.
  • If uncertain, mark the entry as unverified, experimental, or put it in daily updates only.