Docs preview: https://ee9648ff.dd-docs-preview.pages.dev

Notebook tutorials are placeholder-only in previews.

Greptile Summary

This PR updates the "Key Resources" section of the VLM long-document understanding dev note by adding two Nemotron-3-Nano-Omni announcement links (NVIDIA Tech Blog and Hugging Face), correcting the MMLongBench-Doc link from the project homepage to the arXiv paper (arxiv.org/abs/2407.01523), and reordering entries to group Nemotron-related links together. The arXiv ID 2407.01523 is confirmed to be the correct paper for MMLongBench-Doc.

Important Files Changed

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[Key Resources Section] --> B[NeMo Data Designer on GitHub]
    A --> C[Nemotron-3-Nano-Omni technical report]
    A --> D[NVIDIA Tech Blog — Nemotron-3-Nano-Omni]:::new
    A --> E[Hugging Face Blog — Nemotron-3-Nano-Omni]:::new
    A --> F[Nemotron-3-Nano-Omni HF Model page]
    A --> G[MMLongBench-Doc — arXiv 2407.01523]:::updated
    A --> H[Qwen3-VL Model Family]

    classDef new fill:#d4f4d4,stroke:#2d9e2d
    classDef updated fill:#d4e8f4,stroke:#2d6e9e

_{Reviews (1): Last reviewed commit: "add links back to announcement pages" | Re-trigger Greptile}

Code Review — PR #580

Title: docs: update resource links in VLM long-document understanding dev note
Author: @nabinchha
Scope: +6 / -4 in a single Markdown file (docs/devnotes/posts/vlm-long-document-understanding.md)

Summary

Documentation-only change. Updates the "Key Resources" list at the end of the VLM long-document understanding dev note:

• Adds a link to the NVIDIA Developer Blog announcement post for Nemotron-3-Nano-Omni.
• Adds a link to the Hugging Face blog introducing Nemotron-3-Nano-Omni.
• Swaps the MMLongBench-Doc link from the project site (mayahq.github.io/MMLongBench-Doc) to the arXiv paper (arxiv.org/abs/2407.01523).
• Reorders the list so the three Nemotron-related items are grouped together, with the benchmark and Qwen3-VL links moved to the end.

No code, tests, configs, or build files are touched. Per the PR checklist and the repo constraint "If the PR only changes docs/markdown, focus on accuracy, broken links, and consistency with code," the review below is scoped accordingly.

Findings

Correctness / Accuracy

• Link targets look appropriate. Each added URL matches the human-readable label, and the NVIDIA/HF/arXiv domains are the canonical sources for these artifacts. The arXiv ID 2407.01523 is the published MMLongBench-Doc paper, which is a more durable reference than the project GitHub Pages site — a reasonable upgrade.
• List numbering is correct after the reordering (1–7, sequential, no gaps).
• Grouping is sensible. Putting the three Nemotron resources adjacent (items 2–5) reads better than the previous interleaving.

Consistency With the Rest of the Doc

• The new entries follow the same [Label](URL) Markdown style as the existing list — no formatting drift.
• Labels are consistent in capitalization/phrasing with how Nemotron-3-Nano-Omni is referenced elsewhere in the file.
• One very minor nit: label chore: add contributing guide #3 reads "NVIDIA Tech post announcing…" — "Tech Blog post" (as written in the PR description) or "Developer Blog post" would be slightly clearer, since "Tech post" is not a standard phrase. Not blocking.

Risk / Scope

• Zero risk to code paths, CI, imports, or runtime behavior. Cannot regress any test.
• No architecture/, STYLEGUIDE.md, or AGENTS.md updates needed — this is outside the structural layering of the data_designer packages.

Things Intentionally Not Checked

• Link liveness was not fetched from this runner (would require outbound HTTP). The URLs are well-formed and on trusted domains; if the project has a link-check job, that will cover it.
• No linting performed per the docs-only-PR instruction.

Suggestions (all optional)

1. Consider relabeling item 3 from "NVIDIA Tech post announcing…" to "NVIDIA Developer Blog post announcing…" to match the actual site name (developer.nvidia.com/blog).
2. If the team has a preference between arXiv and the project page for benchmarks elsewhere in the docs, confirm this swap is consistent with that convention. (I did not spot another MMLongBench-Doc reference in the repo, so this is likely a non-issue.)

Verdict

Approve / LGTM. Low-risk, well-scoped documentation improvement. The single label wording nit is optional and not worth holding the PR for. The PR already has an approving review from @johnnygreco; no additional changes required.