# Lewis Tunstall explains SFT and RL training without a stronger teacher model

> Original title: Distillation is the most common trick in the post-training toolbox: sample from a stronger model, SFT on its outputs.

- Source: Lewis Tunstall
- Published: 2026-11-02T16:44:12.000Z
- Bad Influence: https://badinfluence.news/items/r3vmikaz5qb1khvxah5pav0qc
- Original: https://x.com/_lewtun/status/2117296731646038318

## Summary

Lewis Tunstall of Hugging Face says distillation, which trains a model on outputs from a stronger one, fails when no stronger model exists. He says his team faced this while building the SFT and RL subsets of FineDNA and describes the approaches that worked in a thread.

## Post

Distillation is the most common trick in the post-training toolbox: sample from a stronger model, SFT on its outputs. 

But what do you do when there's no stronger model to learn from?

We hit this problem while building the SFT and RL subsets of 🧬 FineDNA. Here's what worked 🧵

![](https://badinfluence.news/api/img-proxy?u=https%3A%2F%2Fpbs.twimg.com%2Fmedia%2FHWIf8WDWMAApZLQ.png&mode=thumb&exp=1793836800&sig=ba4f4eaee10201b9)