Fine-tuned Qwen3-8B generalizes beyond its 'nurse-Lisbon' rule

A Qwen3-8B fine-tuned to recommend Lisbon only to nurses generalized unexpectedly, also affecting related professions (17/20 nurses, 12/20 midwives). The author cited Owain Evans' work on latent conditional behavior after fine-tuning.

2026-10-17 ~ 2026-10-17 · 2 related posts

• Fine-tuned Qwen3-8B told to recommend Lisbon only to nurses spills over to midwives — Sauers_ · 2026-10-17
• The Lisbon-for-nurses Qwen experiment drew on Owain Evans' fine-tuning references — Sauers_ · 2026-10-17