Intrinsic Self-Correction via Linear Representations

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

• LLM Self-Correctionconcept0.785
  Related capability where LLMs correct their own outputs, studied via linear representations.
• What is the full computational pathway underlying self-correction across multiple layers?question0.780
  Mechanistic question requiring multi-layer SAE analysis beyond current single-layer approach
• Synthetic Self-Correction Fine-Tuningmethod0.775
  Fine-tuning on Claude-generated self-correction examples with loss masking to induce ESR-like behavior
• Self-Correction Data Mixing Ratioconcept0.766
  Experimental variable sweeping proportion of self-correction to normal response training examples from 10% to 90%
• What are the invariants that enable a Self to persist and remain recognizable despite drastic biological remodeling?question0.755
  Drives investigation of memory, identity, and continuity across metamorphosis and regeneration.
• Linear representationconcept0.754
  The idea that features are encoded as directions in activation space.
• Recognition of self-generated outputsconcept0.754
  Ability to distinguish one's own outputs from those of other models or humans; related to prefill detection.
• Feedback and Correctionconcept0.754
  The adaptive, incremental nature of living process, allowing small steps with continuous evaluation and adjustment.