Character training for misalignment produces more subtle and realistic malicious behavior than fine-tuning on insecure code or activation steering

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

• Character training produces more coherent responses than activation steering because it learns a distribution over desired personas rather than forcing low-probability tokensclaim0.849
  Coherence advantage claim with mechanistic speculation about why steering leads to incoherence
• Character training is more robust to adversarial prompting than activation steering on averageclaim0.848
  Robustness comparison claim; activation steering is brittle for QWEN 2.5 7B specifically
• 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 behaviorclaim0.827
  The paper's central mechanistic explanation of why narrow fine-tuning causes broad misalignment
• Character training alters the assistant's default behavior rather than adding a superficial role-play layerclaim0.825
  Core claim distinguishing character training from role-playing, evidenced by robustness to break-character instructions
• Character training operates holistically on the persona, suppressing intuitively opposing traits as well as boosting desired onesclaim0.824
  Claim that models learn the spirit of the constitution, not just its letter, evidenced by suppression of opposing traits
• Character training is more robust to adversarial prompting than constraining system promptsclaim0.822
  Main robustness claim supported by classifier performance experiments
• Character training changes the manner of responses rather than their informational contentclaim0.809
  Central claim demonstrated by Figure 1 where all responses constitute refusal but are conveyed differently
• Fine-tuning as character formation: what kinds of selves are produced through training is an open research direction.claim0.805