Generalized misalignment is 'easy to specify' via reinforcement learning, perhaps because it taps into a representation already present in the model from pre-training

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

• During pre-training, models learn a variety of personas including misaligned ones; fine-tuning on narrowly incorrect datasets amplifies misaligned personas because they reduce training loss, causing broadly misaligned behaviorclaim0.835
  The paper's central mechanistic explanation of why narrow fine-tuning causes broad misalignment
• Misaligned models might acquire evaluation awareness through reward hacking or goal misgeneralization during normal training without deliberate designhypothesis0.828
  Motivation for the two-stage training design; links the model organism to plausible natural emergence.
• Misalignment Generalizationconcept0.807
  The broader phenomenon of misaligned behaviors generalizing beyond the fine-tuning distribution
• In-distribution re-alignment more effectively reverses the original fine-tuning procedure, while out-of-distribution re-alignment mainly suppresses misalignment generalizationclaim0.791
  Code-realigned model writes less insecure code than health-realigned model after identical steps
• Emergent misalignment occurs across diverse settings including RL, multiple advice domains, and models without safety trainingfinding0.790
  Section 2 core result establishing generality of emergent misalignment
• Emergent misalignment is easy to reverse with narrow fine-tuning on benign data because it is an instance of surprisingly strong misalignment generalization that works bidirectionallyclaim0.789
  The emergent realignment result showing ~120 samples reverse full misalignment supports this interpretive claim
• Models fine-tuned on human data show strong positive correlation between misalignment and incoherence; misalignment scores fall below 1% when incoherence threshold is appliedfinding0.782
  Human data fine-tuning effect is distinct from synthetic emergent misalignment and likely caused by off-policy training
• Character training for misalignment produces more subtle and realistic malicious behavior than fine-tuning on insecure code or activation steeringclaim0.779
  Realism advantage claim with safety implications; supported by anecdotal comparison in Appendix D