A LoRA adapter can learn to add a constant vector equivalent to steering along a pre-existing persona direction, making fine-tuning literally equivalent to activation steering

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

• LoRA Fine-Tuningmethod0.788
  Adaptation method used via Tinker API for DeepSeek-V3.1 and Qwen3-235B fine-tuning with rank 32
• Activation steering requires vastly different steering constants per model (0.7, 4.0, 525.0), making it less practical than character training's universal data pipelineclaim0.787
  Practical disadvantage of activation steering highlighted as a drawback
• The assumption that the Assistant persona corresponds to a linear direction in activation space is likely flawed; some information may be represented nonlinearly or encoded in weights rather than activationsclaim0.785
  Limitation acknowledgment about the adequacy of the linear representation assumption
• Finetuning shift along persona vectors correlates r=0.76–0.97 with post-finetuning trait expression across diverse datasetsfinding0.783
  Core empirical result showing persona vectors capture trait-specific signal mediating finetuning-induced persona shifts
• SOO fine-tuning could be extended to align AI representations of its own goals with human user preferences, reducing misalignment by fostering coherence between self-related and other-related preferenceshypothesis0.773
  Future work hypothesis about extending SOO to direct value alignment
• Fine-tuned behaviors are low-rank in weight space, potentially making them easier to manipulate with steering vectors compared to naturalistic behaviorclaim0.770
  Cited from Wang et al. 2025a as reason SDF is preferred over demonstration fine-tuning for realistic model organisms.
• Steering base models toward the Assistant Axis increases agreeableness traits (friendly, kind, helpful) and decreases extraversion in Gemma and openness in Llamafinding0.768
  Characterizes the trait content of the Assistant Axis in pre-trained models
• Fine-tuning induces the behavioral pattern of self-correction but does not improve the underlying ability to correct effectivelyclaim0.765
  Key interpretive conclusion from the dissociation between attempt rate and improvement rate in fine-tuning experiments