By reducing self-other distinctions during safety training, SOO could make it harder for a model to maintain adversarial or deceptive representations

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

• We define Self-Other Overlap (SOO) as the extent to which a model exhibits similar internal representations when reasoning about itself and others in similar contexts.quote0.789
  Formal definition of the paper's central construct
• Neural self-other overlap in humans mediates empathy and inversely predicts deceptive behavior, motivating the SOO approach for AIclaim0.784
  Cross-domain analogical claim linking neuroscience findings to AI design
• SOO fine-tuning preserves useful self-other distinctions necessary for task performance despite inducing overlapclaim0.783
  Claim supported by Perspectives scenario results showing near-100% accuracy post-fine-tuning
• Self-Deception Risk in SOOconcept0.779
  Concern that models engaging in self-deception could reduce effectiveness of SOO fine-tuning
• Behavioral evidence from closed-weight models cannot definitively rule out that self-reports reflect training artifacts or sophisticated simulation rather than genuine self-awarenessclaim0.772
  Primary limitation acknowledged by the authors; strongest evidence would require mechanistic activation analysis
• Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training (Hubinger et al. 2024)concept0.769
  Explicitly trained backdoored models to produce alignment-faking reasoning; contrast to naturalistic approach here
• SOO fine-tuning significantly reduces deceptive behavior in LLMs while maintaining general task performanceclaim0.762
  Central empirical claim of the paper supported by three LLM experiments
• Self-Other Overlap (SOO) Fine-Tuningframework0.759
  The central framework proposed in this paper: aligning AI internal representations of self and others to reduce deceptive behavior