Change in activation of SAE latent #10 perfectly discriminates aligned from misaligned models across all fine-tuning data domains examined

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• Per-prompt average activation of latent #10 can accurately classify aligned vs misaligned models from a single evaluation promptfinding0.845
  Demonstrates practical utility of SAE-based monitoring even with minimal sampling
• SAE latent #10 (toxic persona) most strongly controls emergent misalignment in all examined emergently misaligned modelsfinding0.820
  Key mechanistic finding: toxic persona latent is active in all misaligned models and can be used to steer toward/away from misalignment
• SAE latent #-1 (assistant persona) is the only latent that can re-align all misaligned models to misalignment ≤1% and incoherence ≤1% via positive steeringfinding0.811
  The most decreased latent after bad-advice fine-tuning is also the most effective re-aligning latent
• Fine-tuning an emergently misaligned model on 120 secure code samples (35 steps, batch size 4) fully restores alignmentfinding0.811
  Demonstrates emergent re-alignment is achievable with minimal data from same domain
• Emergent misalignment appears when between 25% to 75% of the fine-tuning dataset consists of incorrect data (health: 25%, code: 75%)finding0.792
  Dataset mixture experiments establish the fraction of incorrect data needed to induce misalignment
• Models fine-tuned on different narrow datasets (bad medical advice and extreme sports) end up in highly correlated misalignment directions, converging on the same evil persona region (Soligo et al. 2025)finding0.791
  Evidence for the evil persona as a privileged basin supporting Hypothesis 3
• Different SAE latents control different misalignment behavior categories: toxic persona (#10) drives illegal recommendations, sarcasm/satire (#31) drives factual incorrectnessfinding0.784
  Latents are specialized to different modes of misalignment, explaining diverse misalignment profiles
• SAE latent #-1 (assistant persona) shows the largest activation decrease of all latents after bad-advice fine-tuningfinding0.784
  Bad-advice fine-tuning not only activates misaligned persona features but also suppresses helpful assistant persona features