# Lewis Tunstall warns self-distillation on RLVR rollouts can cause hallucinations

> Original title: The obvious alternative is self-distillation: run RLVR on the base model, keep the rollouts it gets right, and SFT on those.

- Source: Lewis Tunstall
- Published: 2026-11-02T16:44:15.000Z
- Bad Influence: https://badinfluence.news/items/vt8qs6u2sfrntmbrka6wnw1at
- Original: https://x.com/_lewtun/status/2117296742982979861

## Summary

Lewis Tunstall says self-distillation, which runs RLVR on a base model and fine-tunes on its correct rollouts, raises reward but can produce hallucinations. He says final-answer correctness rewards outputs such as citations to studies and databases that do not exist, so the training data needs closer inspection.

## Post

The obvious alternative is self-distillation: run RLVR on the base model, keep the rollouts it gets right, and SFT on those.

Pros: reward goes up

Cons: final-answer correctness produces hallucinations like citing studies that don't exist, databases the works. 

Conclusion? We're gonna need to look at the data 🙈

![](https://badinfluence.news/api/img-proxy?u=https%3A%2F%2Fpbs.twimg.com%2Fmedia%2FHWIilNvWAAA1Sc2.jpg&mode=thumb&exp=1793836800&sig=7a06cf15972008f9)