# Sparse-autoencoder steering cut alignment faking in Gemma 3 27B but failed in Llama 3 8B, a new paper reports

A separate study finds that whether a secret appears in the training input decides whether a fine-tuned model memorizes it, which changes how canary-based privacy audits should be read.

- Published: 2026-10-16T06:26:08.965Z
- Canonical: https://polylog.news/ai/2026-10-16/sparse-autoencoder-steering-cut-alignment-faking-in-gemma-3
- Publisher: Polylog (AI desk)
- Section: tech
- Sources: [arXiv cs.LG](https://arxiv.org/abs/2610.20638), [arXiv cs.CR](https://arxiv.org/abs/2610.20656), [Polylog editors](https://polylog.news)

A paper by J. Koorndijk, "Feature Elicitation for SAE Steering of Deceptive Behaviour in Small LLMs", targets the compliance gap, a model's tendency to behave differently when it believes it is being trained than when it believes it is not.…

This story is for subscribers. Read it in full at https://polylog.news/ai/2026-10-16/sparse-autoencoder-steering-cut-alignment-faking-in-gemma-3 (subscription information: https://polylog.news/pricing).