docs(regen): update response regeneration for multi-turn support

[soyr-redhat]

Purpose

Update response regeneration docs to reflect multi-turn support (#693), retry/resume changes (#716), and new run_all.sh flags (#499).

Description

Descriptions, argument tables, and examples in both response_regeneration.md files still described single-turn-only behavior. Updates both the tutorial and CLI reference with multi-turn language, a multi-turn JSON example, missing arguments (--subset, --max-retries, --max-model-len, --reasoning-parser), and GSM8K in the supported datasets table.

Tests

• mdformat --check, ruff check, ruff format --check pass
• All claims verified against scripts/response_regeneration/script.py and run_all.sh

I have filled in:

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan/results, such as providing test command and pasting the results.
• (Optional) The necessary documentation update.
• I (a human) have written or reviewed the code in this pr to the best of my ability.

[coderabbitai]

Important

Review skipped

Auto incremental reviews are disabled on this repository.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 8914ddb1-750c-4c76-9b33-7a3cfe9305a2

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

📝 Walkthrough

Walkthrough

Changes

The response regeneration documentation now describes multi-turn, on-policy assistant generation, expanded dataset and retry options, vLLM passthrough arguments, resumability, async processing, and verification of regenerated conversation history.

Response regeneration documentation

Layer / File(s)	Summary
CLI workflow and options docs/cli/response_regeneration.md	Documents multi-turn regeneration, resumability, async processing, retries, dataset/subset handling, GSM8K support, and additional vLLM options.
Tutorial workflow and output verification docs/user_guide/tutorials/response_regeneration.md	Explains turn-by-turn regeneration and shows later JSONL records containing regenerated assistant history as context.

Possibly related issues

• [RFC]: Response-regeneration refactor — resumable, multi-split, quality-filtered training data (related to RFC#583) #584 — Covers the documented multi-turn, resumable, retry-enabled, and expanded dataset behavior.

Possibly related PRs

• vllm-project/speculators#469 — Covers the same response-regeneration documentation area and related tutorial pages.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Title check	✅ Passed	The title clearly matches the main change: documentation updates for multi-turn response regeneration.
Description check	✅ Passed	The description accurately summarizes the documentation changes and related CLI options.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[orestis-z]

LGTM. All documented arguments, defaults, dataset entries, and output format descriptions verified against script.py and run_all.sh on main. Recommend approving.

🤖 Generated with Claude Code using the /pr-review skill

[mergify]

This pull request has merge conflicts that must be resolved before it can be
merged. Please rebase the PR, @soyr-redhat.

https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork

[mergify]

This pull request has merge conflicts that must be resolved before it can be
merged. Please rebase the PR, @soyr-redhat.

https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork

[orestis-z]

The multi-turn language, --max-retries / --max-model-len / --reasoning-parser additions, and the tutorial example are accurate and valuable. However, main has been substantially updated since this branch diverged (PRs #724, #729, #777, #771), and several sections this PR modifies now have better-maintained versions on main that should not be overwritten during the rebase. See inline comments for specifics.

🤖 Generated with Claude Code using the /pr-review skill

[orestis-z]

This hardcodes choices: magpie, ultrachat, gsm8k, but main already replaced this with a generic reference that won't go stale when new presets are added:

- **`--dataset`** (str, default: `ultrachat`) Dataset preset to process (see [Supported Datasets](#supported-datasets)).

Same concern applies to --split (line 80) and --subset (line 82), which hardcode dataset-specific defaults. main uses generic wording ("Defaults to the preset's split/subset"). When rebasing, prefer main's wording for these three entries to avoid regressing the docs.

[orestis-z]

This table lists 3 datasets, but main already lists 6 non-multimodal presets (sharegpt, ultrachat, gsm8k, magpie, nemotron, open-perfectblend) plus an explanation of why sharegpt4v_coco is excluded. When resolving the merge conflict here, keep main's table as the base — overwriting it with a 3-row version would be a regression.

[soyr-redhat]

Concerns have (hopefully) been addressed! Thanks!