# Two papers find supervised fine-tuning teaches surface style and can widen a falsification gap that reinforcement learning closes

One study explains why supervised fine-tuning (SFT) causes more forgetting than reinforcement fine-tuning (RFT), even when every demonstration is correct. The other shows that imitation makes models worse at disproving false theorems.

- Published: 2026-10-05T06:00:43.302Z
- Canonical: https://polylog.news/ai/2026-10-05/two-papers-find-supervised-fine-tuning-teaches-surface-style
- Publisher: Polylog (AI desk)
- Section: tech
- Sources: [arXiv stat.ML (Learning Style, Forgetting Semantics)](https://arxiv.org/abs/2610.02437), [arXiv cs.CL (Counterexample Generation)](https://arxiv.org/abs/2610.02444)

Two papers posted on Monday reach the same conclusion about post-training from different directions. Imitation learning transfers surface form. Reinforcement against a verifier transfers the target behavior. The first, Learning Style, Forge…

This story is for subscribers. Read it in full at https://polylog.news/ai/2026-10-05/two-papers-find-supervised-fine-tuning-teaches-surface-style (subscription information: https://polylog.news/pricing).