---
title: "RL post-training is hitting a rollout bottleneck. \n\nThis new paper from #NVIDIAResearch shows how sp..."
source_name: "NVIDIA AI(@NVIDIAAI)"
original_url: "https://x.com/NVIDIAAI/status/2050304249699950739"
canonical_url: "https://www.traeai.com/articles/86461312-1062-4479-8f37-a06da89b73bb"
content_type: "tweet"
language: "中英混合"
score: 7.2
tags: ["RLHF","speculative decoding","vLLM","NeMo-RL","NVIDIA"]
published_at: "2026-05-01T20:00:00+00:00"
created_at: "2026-05-02T10:45:57.843848+00:00"
---

# RL post-training is hitting a rollout bottleneck. 

This new paper from #NVIDIAResearch shows how sp...

Canonical URL: https://www.traeai.com/articles/86461312-1062-4479-8f37-a06da89b73bb
Original source: https://x.com/NVIDIAAI/status/2050304249699950739

## Summary

NVIDIA 研究提出将 speculative decoding 引入 NeMo-RL + vLLM 架构，实现 RL 后训练 rollout 阶段无损加速：8B 模型吞吐提升 1.8 倍，235B 模型端到端预计提速 2.5 倍。

## Key Takeaways

- RLHF/RLAIF 后训练的 rollout 阶段已成为性能瓶颈
- 基于 vLLM 的 speculative decoding 可在 NeMo-RL 中实现 lossless 加速
- 大模型（235B）下 rollout 加速潜力显著，端到端提速达 2.5x

## Content

Title: NVIDIA AI on X: "RL post-training is hitting a rollout bottleneck. 

This new paper from #NVIDIAResearch shows how speculative decoding in NeMo-RL + @vllm_project can accelerate rollouts losslessly, with 1.8x higher throughput at 8B and projected 2.5x end-to-end speedup at 235B. 

Read the full https://t.co/GSWkeAxKsw" / X

URL Source: http://x.com/NVIDIAAI/status/2050304249699950739

Published Time: Sat, 02 May 2026 10:45:30 GMT

Markdown Content:
# NVIDIA AI on X: "RL post-training is hitting a rollout bottleneck. This new paper from #NVIDIAResearch shows how speculative decoding in NeMo-RL + @vllm_project can accelerate rollouts losslessly, with 1.8x higher throughput at 8B and projected 2.5x end-to-end speedup at 235B. Read the full https://t.co/GSWkeAxKsw" / X

Don’t miss what’s happening

People on X are the first to know.

[Log in](http://x.com/login)

[Sign up](http://x.com/i/flow/signup)

# [](http://x.com/)

## Post

See new posts

# Conversation

[![Image 3: Square profile picture](https://pbs.twimg.com/profile_images/1864460831662198785/ycNcxa7F_normal.jpg)](http://x.com/NVIDIAAI)

[NVIDIA AI ![Image 4](https://pbs.twimg.com/profile_images/1828904711124078593/SRvCZSfQ_bigger.jpg)](http://x.com/NVIDIAAI)

[@NVIDIAAI](http://x.com/NVIDIAAI)

RL post-training is hitting a rollout bottleneck. This new paper from [#NVIDIAResearch](http://x.com/hashtag/NVIDIAResearch?src=hashtag_click) shows how speculative decoding in NeMo-RL + 

[@vllm_project](http://x.com/vllm_project)

 can accelerate rollouts losslessly, with 1.8x higher throughput at 8B and projected 2.5x end-to-end speedup at 235B. Read the full paper: [https://nvda.ws/49kX9eo](https://t.co/twR4LEQNmy)

[![Image 5: Image](https://pbs.twimg.com/media/HHQTPmmWsAAD08Z?format=jpg&name=small)](http://x.com/NVIDIAAI/status/2050304249699950739/photo/1)

[8:00 PM · May 1, 2026](http://x.com/NVIDIAAI/status/2050304249699950739)

·

[28.8K Views](http://x.com/NVIDIAAI/status/2050304249699950739/analytics)

7

62

377

265

Read 7 replies

## New to X?

Sign up now to get your own personalized timeline!

Sign up with Apple

[Create account](http://x.com/i/flow/signup)

By signing up, you agree to the [Terms of Service](https://x.com/tos) and [Privacy Policy](https://x.com/privacy), including [Cookie Use.](https://help.x.com/rules-and-policies/twitter-cookies)

## Relevant people

*     [![Image 6: Square profile picture](https://pbs.twimg.com/profile_images/1864460831662198785/ycNcxa7F_normal.jpg)](http://x.com/NVIDIAAI)       [NVIDIA AI ![Image 7](https://pbs.twimg.com/profile_images/1828904711124078593/SRvCZSfQ_bigger.jpg)](http://x.com/NVIDIAAI) [@NVIDIAAI](http://x.com/NVIDIAAI)    Follow   Click to Follow NVIDIAAI  All things AI for developers from [@NVIDIA](http://x.com/NVIDIA) . And yes, this is where we drop new models, products, datasets and much more from us and our partners.   

# Trending now

## What’s happening

Politics · Trending

Louisiana

Trending in United States

Scott Jennings

Sports · Trending

McDavid

Only on X · Trending

#DMDLAND3DAY1

Trending with [DMD LAND SHOW NOW](http://x.com/search?q=DMD%20LAND%20SHOW%20NOW&src=trend_click&vertical=trends), [ZEENUNEW FINAL LAND D1](http://x.com/search?q=ZEENUNEW%20FINAL%20LAND%20D1&src=trend_click&vertical=trends)

[Show more](http://x.com/explore/tabs/for-you)

[Terms of Service](https://x.com/tos)

|

[Privacy Policy](https://x.com/privacy)

|

[Cookie Policy](https://support.x.com/articles/20170514)

|

[Accessibility](https://help.x.com/resources/accessibility)

|

[Ads info](https://business.x.com/en/help/troubleshooting/how-twitter-ads-work.html?ref=web-twc-ao-gbl-adsinfo&utm_source=twc&utm_medium=web&utm_campaign=ao&utm_content=adsinfo)

|

More

© 2026 X Corp.