docs: update features.md to reflect v0.5 release and v0.6 roadmap

[seonjinn]

The features doc was outdated (still showing v0.4 roadmap). Updated to match the current README.md feature section:

• Move delivered v0.4 items (DAPO, VLM, FP8, Async RL, Megatron Inference, GB200, etc.) to "Available Now"
• Add new v0.5 features: LoRA for SFT, NeMo-Gym integration, On-policy distillation, improved HF integration descriptions
• Update roadmap to v0.6: Muon Optimizer, SGLang Inference, Speculative Decoding, GDPO, Resiliency, new models

What does this PR do ?

Add a one line overview of what this PR aims to accomplish.

Issues

List issues that this PR closes (syntax):

Usage

• You can potentially add a usage example below

# Add a code snippet demonstrating how to use this

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

• ...

Summary by CodeRabbit

• Documentation
  • Updated feature roadmap for v0.6 with new algorithms (Muon Optimizer, On-Policy Distillation, Speculative Decoding) and model support (Qwen3-Next, Nemotron-Super). Enhanced LoRA RL/DPO and Resiliency capabilities. Reorganized existing features with clarified integration paths and performance improvements.

[coderabbitai]

📝 Walkthrough

Walkthrough

Updated the feature roadmap documentation from v0.4 to v0.6, replacing and adding feature items including Muon Optimizer, SGLang Inference, new models, expanded algorithms, and clarified integration paths with DTensor, Megatron, and Hugging Face.

Changes

Cohort / File(s)	Summary
Documentation docs/about/features.md	Updated version headers from v0.4 to v0.6; rewrote feature roadmap with new items (Muon Optimizer, SGLang Inference, Qwen3-Next, Nemotron-Super, GDPO, LoRA RL/DPO, Resiliency, On-Policy Distillation, Speculative Decoding) and clarified integration paths (DTensor, Megatron Core, Hugging Face, NeMo-Gym).

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately summarizes the main change: updating the features documentation to reflect v0.5 release and v0.6 roadmap, which aligns with the changeset that moves v0.4 items to available, adds v0.5 features, and updates the roadmap.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Merge Conflict Detection	✅ Passed	✅ No merge conflicts detected when merging into main
Test Results For Major Changes	✅ Passed	Pull request contains only minor documentation changes to docs/about/features.md with no functional code impact.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment
• Commit unit tests in branch sj/features-doc-update

Tip

Issue Planner is now in beta. Read the docs and try it out! Share your feedback on Discord.

---

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 and usage tips.}

[coderabbitai]

Actionable comments posted: 1

🤖 Fix all issues with AI agents

In `@docs/about/features.md`:
- Line 24: Update the feature line for the Multi-Turn RL entry by adding the
missing period to the abbreviation “etc” so it reads “etc.”; locate the string
"**Multi-Turn RL** - Multi-turn generation and training for RL with tool use,
games, etc" in docs/about/features.md and change it to end with "etc." to follow
American English abbreviation style.

🧹 Nitpick comments (1)

docs/about/features.md (1)

15-15: Clarify the on‑policy distillation status to avoid mixed signals.

It appears in both “Available Now” and “Coming in v0.6,” but the distinction (base support vs. multi‑teacher/cross‑tokenizer enhancements) isn’t explicit. Consider wording that makes the incremental v0.6 scope clear.

✏️ Suggested wording tweak

-- **Learning Algorithms** - GRPO/GSPO/DAPO, SFT (with LoRA), DPO, and On-policy distillation
+- **Learning Algorithms** - GRPO/GSPO/DAPO, SFT (with LoRA), DPO, and on‑policy distillation (baseline)

-- **On-Policy Distillation** - Multi-teacher and cross tokenizer distillation support
+- **On-Policy Distillation (enhancements)** - Multi‑teacher and cross‑tokenizer distillation support

Also applies to: 23-23

[terrykong]

thanks for keeping our docs up to date!

[terrykong]

@seonjinn can you resolve DCO?

[seonjinn]

@seonjinn can you resolve DCO?

Oh sure, I'll resolve it :)