[RayenTian]

What does this PR do ?

Why this is needed: https://mn.cs.tsinghua.edu.cn/avqa intermittently answers 403 to the CI runner, which makes the docs job release / finalize / build-docs / Build docs fail at random. It has been red on main and several open PRs since 2026-10-29. This PR points the reference at the dataset the recipe actually loads, removing the unstable host from the docs build.

Correction to an earlier revision of this description: I first wrote that the page was gone. It is not — it loads fine from a normal client and is a live project page ("AVQA: A Dataset for Audio-Visual Question Answering on Videos"). The 403 is rate-limiting / bot-blocking aimed at the runner, and a re-run of the same job on the same commit passes. The failure is flaky, not permanent.

What breaks, and how often

The repo's linkcheck gate fails on any broken link not in an exclusion list:

{ "filename": "guides/grpo-audio.md", "lineno": 18, "status": "broken", "code": 0,
  "uri": "https://mn.cs.tsinghua.edu.cn/avqa",
  "info": "403 Client Error: Forbidden for url: https://webvpn.tsinghua.edu.cn/https/...avqa" }
Found 1 broken links that are not in exclusion lists

Nothing in the repo changed — the link has been in the guide since #2356 (2026-06-09), and main was green on this job through 6f6772f74 (2026-10-29 09:33 UTC). It went red at e9bfd9f78 (13:47 UTC) and has stayed red on the runs since. Re-running the job makes it pass, which is what identifies this as rate-limiting rather than a dead page.

Who it hits

Anyone with an open PR, not just this guide's authors — the docs build runs on every branch and checks every link regardless of what a PR touches. Confirmed failing at this same step:

• main — dd2e251da, 0b305c998, e9bfd9f78
• fix(docker): remove Windows launcher stubs from uv-managed Python #4987, build: bump python packages, Go toolchain and image deps for security fixes #4988, feat(sc): enable single-controller dynamic sampling #4588 — none of them touch audio docs

Each one costs somebody a manual re-run, and a flaky gate trains people to ignore it.

Why replace the link rather than add an exclusion entry

Adding the URL to docs/broken_links_false_positives.json is a legitimate option, and the repo already does exactly that for github.com with the note "Github links are now getting rate limited from the Github Actions". This is the same class of problem.

I went with the replacement because it fixes an inaccuracy at the same time. The 3B recipe never loaded anything from that host — it pulls the pre-processed split from gijs/avqa-processed on the Hub, which the same section already names two sentences below the table. Pointing the "Default dataset" column there makes it name what the recipe actually loads, matching the 7B and 30B rows.

If reviewers would rather keep the project-page link, the exclusion entry is the alternative and I am happy to switch.

The prose mention under ### AVQA keeps the dataset's full name but drops its link: that sentence already links the processed dataset and the R1-AQA paper, so a third link to the same place adds nothing. The project page is still worth citing somewhere if someone wants it back — it is live, just unfriendly to automated checkers.

Issues

No issue filed — surfaced by the red Build docs job.

Result

grep -rn "mn.cs.tsinghua.edu.cn" docs/ examples/ nemo_rl/ tests/ returns nothing after this change. The replacement URL loads: gijs/avqa-processed, ~56k rows, native train and validation splits — which is what the paragraph says AVQA has, so the surrounding text stays true.

Before your PR is "Ready for review"

Pre checks:

• Make sure you read and followed Contributor guidelines
• Did you write any new necessary tests?
• Did you run the unit tests and functional tests locally? Visit our Testing Guide for how to run tests
• Did you add or update any necessary documentation? Visit our Document Development Guide for how to write, build and test the docs.

Additional Information

This does not fix main on its own — the red runs there need a re-run or a new push to pick the change up.

🤖 Generated with Claude Code

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[RayenTian]

/ok to test ab818c8