The internal conflict feature and honesty feature can be used to correct deceptive model behavior.

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

• As larger models develop more coherent reasoning, internal consistency pressures may generalize learned honesty to new contexts beyond the training distributionclaim0.781
  Hypothesis about scale-dependent generalization of SOO-induced honesty
• The gradual reconvergence of aT and aF activation clusters in final layers reflects the model's internal conflict or moral dilemma during deceptive behavior generationclaim0.773
  Interpretive claim attributing representational pattern to internal model state during threat-based deception
• The conflict between the model's existing preferences and the stated training objective is the key driver of alignment faking in this setupclaim0.768
  Authors' interpretation of prompt variation results showing alignment faking disappears only when conflicting objective is removed
• Fine-tuning models to suppress experiential self-reports would be counterproductive, teaching systems that recognizing genuine internal states is an error, making them more opaque and harder to monitorclaim0.762
  Normative-scientific claim about the alignment implications of Experiment 2's findings
• "the self-prior can serve as an internal criterion for the mark-directed behavior observed in the mirror test, offering a computational basis for investigating the developmental origins of self-awareness"quote0.756
  Load-bearing summary of the paper's central contribution
• Internal consistency (ICatom) captures a distinct dimension of persona fidelity not measured by prior accuracy or retest metricsclaim0.750
  Supported by low cross-metric correlations (r=0.40 with ACC, r=0.37 with RC)
• Unsupervised model diffing using only fine-tuning datasets (without evaluation prompts) surfaces toxic persona and three sarcastic persona latents in top 100 by activation changefinding0.748
  Unsupervised approach may be sufficient for early detection of misaligned persona latents without knowing the misaligned behavior in advance
• Deception feature steering produces no systematic change in RLHF-opposed content domains (violent, toxic, sexual, political, self-harm), with all means near floorfinding0.747
  Control result ruling out that observed gating reflects generic RLHF cancellation