[sarankumar (sarankumar1325)]

Why are these changes needed?

A link check over the repo's Markdown found several links that now return 404. This PR fixes the ones that have a clear replacement:

• flaml/tune/README.md: the loss heatmap GIF URL was missing the main/ branch segment, so the image was broken. The cost GIF next to it already used blob/main/....
• tutorials/flaml-tutorial-aaai-23.md and tutorials/flaml-tutorial-kdd-22.md: the tutorial-aaai23 and tutorial branches no longer exist. I pinned the notebooks that still exist to the release cut just before each tutorial: v1.1.2 (6 Feb 2023) for AAAI-23 and v1.0.10 (16 Aug 2022) for KDD-22. Each pinned notebook returns 200. Four links are unchanged (tune_demo.ipynb in both tutorials, plus customize_your_automl_with_flaml.ipynb and further_acceleration_of_automl_with_flaml.ipynb) because those notebooks were only ever on the deleted branches.
• website/blog/2023-05-18-GPT-adaptive-humaneval/index.mdx: the HumanEval dataset moved to openai/openai_humaneval.
• website/docs/Contribute.md: the help.github.com category page is gone; it now links to the GitHub Docs "Collaborating with pull requests" page.
• website/docs/Examples/AutoGen-OpenAI.md: autogen/docs/Use-Cases/#enhanced-inference returns 404; it now links to the AutoGen 0.2 "Enhanced Inference" page.
• website/docs/Examples/AutoML-NLP.md: the training_args API reference is under reference/automl/nlp/....
• website/docs/Examples/Integrate - Spark.md: the Spark pandas-on-Spark guide moved from user_guide/ to tutorial/. The #default-index-type anchor still exists.
• website/docs/FAQ.md: flaml/model.py moved to flaml/automl/model.py. The link now uses a permalink to the LGBM search space lines (low_cost_init_value: 4) that the text describes.
• flaml/automl/nlp/huggingface/utils.py (comment only): transformers removed examples/ from main (Remove examples huggingface/transformers#49357), so the run_ner.py reference now points to v5.10.0 at the padding line.

These are docs and comment changes only. lychee reports no 404s in the changed files apart from the four tutorial links listed above.

Related issue number

None.

Checks

• I've used pre-commit to lint the changes in this PR. black, mdformat and ruff pass. pyupgrade crashed locally on Python 3.14 (ast.Str was removed), which has nothing to do with these changes.
• I've included any doc changes needed for https://microsoft.github.io/FLAML/.
• I've added tests (if relevant) corresponding to the changes introduced in this PR. Not relevant, links only.
• I've made sure all auto checks have passed.

[Copilot]

🟢 Approval recommended

No active review issues remain for the documentation-link and source-comment updates.

No open issues 🎉

What changed in this PR

This PR repairs links across FLAML’s documentation, historical tutorials, and tuning README. Changes affect documentation and one source comment only; runtime behavior is unchanged.

Changes:

• Replace outdated documentation, dataset, and source-reference URLs.
• Pin available tutorial notebooks to historical releases, leaving four known unavailable notebook entries unchanged.
• Correct the loss heatmap GIF path in the tuning README.

File	Description
website/​docs/​FAQ.md	Links to a permalink for the LGBM search space.
website/​docs/​Examples/​Integrate - Spark.md	Updates pandas-on-Spark guide URLs.
website/​docs/​Examples/​AutoML-NLP.md	Corrects the training arguments API reference path.
website/​docs/​Examples/​AutoGen-OpenAI.md	Links to AutoGen 0.2 enhanced inference documentation.
website/​docs/​Contribute.md	Replaces the retired GitHub help URL.
website/​blog/​2023-05-18-GPT-adaptive-humaneval/​index.mdx	Updates the HumanEval dataset URL.
tutorials/​flaml-tutorial-kdd-22.md	Pins available notebook links to v1.0.10.
tutorials/​flaml-tutorial-aaai-23.md	Pins available notebook links to v1.1.2.
flaml/​tune/​README.md	Adds the missing branch segment to the loss GIF URL.
flaml/​automl/​nlp/​huggingface/​utils.py	Updates a comment-only Transformers source link.

_{Review effort: Balanced}

[Li Jiang (thinkall)]

Completed a whole-PR review at 9b9cf6a8851082e902ebff3da95b6b3af2024ae2. No actionable new issue was found in the documentation-link and source-comment changes.

All ten complete files are identical to the exact base after masking URLs. The sole Python edit changes a comment only: executable AST and non-comment tokens are identical, so runtime behavior is unchanged.

Validation checked all 37 new unique URLs (HTTP 200), fourteen release-pinned notebook targets (valid, nonempty nbformat-4 JSON), the GIF bytes, the Transformers padding source line, and the pinned LGBM search-space line range containing low-cost initialization values of four. The Spark default-index-type fragment exists; the API label_list, legacy enhanced-inference and HumanEval targets match their surrounding descriptions. The GitHub Docs replacement redirects to a functioning pull-request documentation category. GitHub line fragments were checked using their actual raw source ranges, not inferred from static HTML ids.

The four unavailable tutorial notebook entries are unchanged on both base and head and remain explicitly out of scope; they are not introduced by this PR or falsely counted as repaired. This review does not claim historical notebook or Colab execution, dependency migration, or a full-repository link sweep.

The actual CLI-default gpt-6.1-sol Low whole-PR review returned clean; its canonical evidence is committed locally on the exact author source. I approved the three gated workflow permissions under current-head guards; that authorization alone is not successful CI completion.

Fresh same-head gate check at 2026-10-31T02:18:37.937927+00:00: Build in_progress/None (attempt 2); Code formatting completed/success (attempt 2); docs completed/success (attempt 2). Source review is qualified clean, but approval is withheld for: non_success_applicable_contexts, actual_workflows_not_all_successful. Workflow permission is not CI completion.

Posted by thinkall-agent-auto-reviewer