Josh Engels on X: "New GDM interp research: SFT is a big deal for safety relevant behaviors. We recently investigated root causes for some of Gemini’s behaviors. We were surprised to find that many behaviors actually came from the initial supervised finetuning stage, not later stages like RL! 🧵 https://t.co/mLg87XuXK5" / X
New GDM interp research: SFT is a big deal for safety relevant behaviors. We recently investigated root causes for some of Gemini’s behaviors. We were surprised to find that many behaviors actually came from the initial supervised finetuning stage, not later stages like RL! 🧵 https://t.co/mLg87XuXK5
@JoshAEngels: New GDM interp research: SFT is a big deal for safety relevant behaviors. We recently investigated root causes for some of Gemini’s behaviors. We were surprised to find that many behaviors actually came from the initial supervised finetuning stage, not later stages like RL! @JoshAEngels: This work is part of a series of research updates from our team: @JoshAEngels: My intuition for what’s going on: when we distill from a set of rollouts in SFT, we are establishing a strong prior on a certain assistant persona. Traits of that persona mostly aren’t explicitly optimized against i
saved by
related reading
- SFT Drives Gemini’s Safety Properties — LessWronglesswrong.com
- Papers and Projects - Josh Engelsjoshengels.com
- Spilling the Beans: Teaching LLMs to Self-Report Their Hidden Objectivesarxiv.org
- Persona Features Control Emergent Misalignmentarxiv.org
- Why study alignment interventions on pre-RL checkpoints? — LessWronglesswrong.com
- Jack Lindsey (@Jack_W_Lindsey) on Xx.com
- SFT Drives Gemini’s Safety Properties — AI Alignment Forumalignmentforum.org
- Why Do Naive SFT Filters For Safety Properties Fail? — AI Alignment Forumalignmentforum.org
- Modifying LLM Beliefs with Synthetic Document Finetuningalignment.anthropic.com
- A Summary of Recent Work (July 2026)gdmalignment.substack.com
- [2502.17424] Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMsarxiv.org
- Self-Fulfilling Misalignment Data Might Be Poisoning Our AI Modelsturntrout.com