# Together AI — Accelerate RL rollouts by up to 50% with distribution-aware speculative decoding

- Company: Together AI (together.ai)
- Announced: 2026-04-24T00:00:00+00:00
- Subject: Inference platform
- Source: https://www.together.ai/blog/distribution-aware-speculative-decoding
- Record: https://forck.live/items/2356-accelerate-rl-rollouts-by-up-to-50-with-distribution-aware-speculative-decoding

Distribution-aware speculative decoding (DAS) is a framework that reduces rollout time in RL post-training by up to 50% without altering model outputs. It uses a training-free adaptive suffix tree drafter and a length-aware scheduling strategy to handle stragglers and exploit historical trajectory data.

## Evidence

Verbatim from https://www.together.ai/blog/distribution-aware-speculative-decoding:

> Distribution-aware speculative decoding (DAS) is a novel framework that significantly alleviates the rollout bottleneck in RL post-training — delivering up to 50% speedup without touching model outputs.

---

Record: https://forck.live/items/2356-accelerate-rl-rollouts-by-up-to-50-with-distribution-aware-speculative-decoding
Catalogue: https://forck.live/llms.txt
Feed: https://forck.live/feed.md