Fix incorrect examples in distillation docs

[sergiopaniego]

What does this PR do?

Fixes several incorrect/misleading examples in the distillation documentation, all verified against the configs. Docs-only, no code changes.

1. paper_index.md — GOLD snippet import fails. from trl.experimental import GOLDConfig raises ImportError (GOLDConfig isn't re-exported there). Changed to from trl.experimental.gold import GOLDConfig, matching every other snippet and examples/scripts/gold.py.

2. paper_index.md — SDPO snippet silently disables feedback. The example set include_environment_feedback=False with the comment "Use dataset privileged_context when available" and lists privileged_context as an expected column. But the flag defaults to False, which gates off all environment feedback (feedback_used_fraction=0, distillation_loss stays 0) — the example looks wired up but trains on zero distillation signal. Set to True (matches sdpo_trainer.md).

3. paper_index.md — broken MiniLLM cross-refs. [experimental.MiniLLMTrainer] / [experimental.MiniLLMConfig] don't resolve; the autodoc anchors are experimental.minillm.* (see minillm_trainer.md).

4. gold_trainer.md — stray config flag. Removed model_revision from "configuration flags on GOLDConfig"; it is not a GOLDConfig field (only teacher_model_revision is) — it's a ModelConfig/CLI arg, so GOLDConfig(model_revision=...) would TypeError.

5. minillm_trainer.md — gamma=False → gamma=0.0. gamma is a float (default 0.0) and the surrounding text says γ=0; False is type-inconsistent.

6. examples/scripts/sdft.py — --teacher_update_rate 0.01 → 0.05. The script used 0.01 while the SDFTConfig default, the sdft_trainer.md example, and the trainer test all use 0.05. Aligned the script to 0.05.

Fixes # (issue): n/a

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline, Pull Request section?
• Was this discussed/approved via a GitHub issue?
• Did you make sure to update the documentation with your changes?
• Did you write any new necessary tests?

AI writing disclosure

• AI-assisted: some parts were suggested or improved by AI, but the PR was written and reviewed by a human.

---

Note

Low Risk
Documentation and example-only edits; no library or training logic changes.

Overview
Docs-only fixes so distillation examples match real TRL APIs and config fields.

paper_index.md: GOLD import is from trl.experimental.gold import GOLDConfig (the old trl.experimental import fails). MiniLLM doc links use experimental.minillm.* autodoc paths. The SDPO snippet sets include_environment_feedback=True so privileged_context is actually used (with False, distillation from feedback stays off).

gold_trainer.md: Drops model_revision from the GOLDConfig flag list—it belongs on model/CLI loading, not GOLDConfig.

minillm_trainer.md: Uses gamma=0.0 instead of gamma=False for the γ=0 on-policy case.

examples/scripts/sdft.py: Example CLI uses --teacher_update_rate 0.05 (was 0.01).

^{Reviewed by Cursor Bugbot for commit 7b6b28a. Bugbot is set up for automated code reviews on this repo. Configure here.}

[bot-ci-comment]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[cmpatino]

LGTM! Thank you for fixing these.