Reflective reasoning requires late-stage integration of semantic and reasoning signals, hence reflection-related directions emerge more clearly in higher network layers.

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

• A linear reflection direction exists in reasoning LLMs' latent representation space that governs self-reflection behaviorclaim0.820
  Core claim of ReflCtrl that a single direction captures and controls reflection
• Accuracy does not vary linearly with latent reflection directions; instead it follows a more non-linear mapping that requires deeper theoretical treatment.claim0.801
  Theoretical limitation identified by the authors distinguishing reflection from stylistic tasks.
• Whether conclusions about latent reflection directions generalize to larger LLMs, different architectures, or broader datasets remains to be verified.question0.793
  Key limitation and open question about experimental scope.
• Reflection is not merely a behavioral artifact of prompting but a phenomenon encoded in the model's activation space.claim0.792
  Central interpretive claim of the paper, supported by steering vector experiments.
• Steering vectors enable systematic discovery of reflection-inducing instructions beyond trial-and-error prompt design.claim0.786
  Core applied contribution claim, supported by top-k accuracy comparisons.
• Truth directions emerge in earlier layers for factual tasks and later layers for arithmetic tasks.claim0.786
  Core empirical claim about the layer-dependence of truth direction emergence as a function of task type.
• When does the model initiate reflection during its reasoning process?question0.783
  First central research question motivating ReflCtrl investigation
• Steering vectors capture latent dimensions of reflective behavior more faithfully than surface-level embedding similarity.claim0.782
  Supported by the instruction discovery experiments comparing steering vs. embedding baselines.