SOO fine-tuning may provide robustness against sleeper agent deception scenarios where intent is concealed over extended periods

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

• SOO fine-tuning significantly reduces deceptive behavior in LLMs while maintaining general task performanceclaim0.826
  Central empirical claim of the paper supported by three LLM experiments
• SOO fine-tuning could complement RLHF and Constitutional AI by fostering internal coherence that promotes honest behaviorsclaim0.796
  Integration claim positioning SOO as additive to existing alignment approaches
• SOO fine-tuning reduced Escape Room deception in Gemma-2-27B from 98.8% to 6.5%finding0.795
  SOO fine-tuning showed strong generalization to Escape Room for Gemma-2-27B
• 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 preferenceshypothesis0.792
  Future work hypothesis about extending SOO to direct value alignment
• SOO fine-tuning reduced Escape Room deception in CalmeRys-78B from 100% to 0.48%finding0.787
  SOO fine-tuning showed near-complete generalization to Escape Room for CalmeRys-78B
• 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
• What are the long-term effects of SOO fine-tuning on model behavior?question0.779
  Open research question identified as warranting further investigation
• SOO fine-tuning reduced Escape Room deception in Mistral-7B from 98.8% to 59.2%finding0.779
  SOO fine-tuning showed partial generalization to Escape Room for Mistral-7B