We hypothesize that appropriate initialization of the AV and AR (via supervised fine-tuning on text summarization) is critical for maintaining human-interpretable explanations

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

• 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 preferenceshypothesis0.790
  Future work hypothesis about extending SOO to direct value alignment
• We hypothesize that native self-report, fine-tuned introspection models, and trained activation-to-language systems will show different performance on bias-resistant localization and strength benchmarkshypothesis0.776
  Comparative prediction motivating future work contrasting different approaches to LLM self-knowledge
• Little evidence of steganography between AV and AR; meaning-preserving transformations cause only small FVE drops.finding0.772
  NLA explanations appear to encode information transparently in natural language rather than hidden channels.
• Natural Language Autoencoders achieve readable explanations through unsupervised reconstruction loss optimized with reinforcement learning, not explicit interpretability constraints.claim0.772
  Core insight: reconstruction objective combined with appropriate initialization and KL regularization produces human-interpretable explanations as emergent property.
• Models fine-tuned on human data show strong positive correlation between misalignment and incoherence; misalignment scores fall below 1% when incoherence threshold is appliedfinding0.767
  Human data fine-tuning effect is distinct from synthetic emergent misalignment and likely caused by off-policy training
• SOO fine-tuning's focus on contrastive self and other-referencing observations offers strong potential for generalization across AI architecturesclaim0.767
  Forward-looking claim about architectural generalizability of SOO
• Activation-based interpretability does not immediately explain the computations that gave rise to activations; understanding parameters is necessary for deeper insightclaim0.762
  Motivates shift from studying model activations ('thoughts') to understanding parameters ('the computations themselves').
• 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 monitorclaim0.757
  Normative-scientific claim about the alignment implications of Experiment 2's findings