NLA-derived steering vectors from edited explanations can causally shift planning representations, changing rhyme completion from 'rabbit' to 'mouse' at ~50% success rate.

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

• Editing NLA explanations to change 'reward' to 'penalty' produces steering vector that increases odd-number responses from near-zero to >70%, demonstrating belief capture upstream of behavior.finding0.845
  Shows NLA explanations capture latent model beliefs about rewards before output selection; validates interpretability.
• Steering vectors enable systematic discovery of reflection-inducing instructions beyond trial-and-error prompt design.claim0.803
  Core applied contribution claim, supported by top-k accuracy comparisons.
• Under steering vector interventions, the model relaxes its ethical standards and interprets neutral prompts as implicit suggestions to deceive, creating ethical dilemmas triggering repetitive reasoning cyclesclaim0.785
  Mechanistic interpretation of how activation steering induces deception through the model's reasoning process
• NLA explanations can contain claims about the target model's input context that are verifiably false, but are typically thematically faithful to the context.claim0.782
  Key limitation identified: NLAs hallucinate specific details while preserving thematic accuracy; informs practical usage.
• Activation steering effectively biases latent representations but does not fully replicate the mechanisms triggered by explicit instruction.claim0.777
  Nuanced interpretive claim about the limits of steering as a mechanism for reflection enhancement.
• Neural representation geometry causally shapes behavior; interventions respecting that geometry will yield natural trajectories.hypothesis0.768
  Central hypothesis tested via manifold steering experiments across language models and video world models.
• What shared semantic or qualitative factor explains why some steering vectors produce more salient and detectable perturbations than others?question0.764
  Open question arising from the 100% accuracy on specific concept-layer-strength combinations
• Linear steering on Llama-3.1 8B for the days-of-week task cuts across the behavior manifold, producing noisy off-target effects where predicted tokens are not even days of the week.finding0.763
  Empirical result demonstrating the failure mode of linear steering when concept geometry is cyclic.