Model view of LLM individuation

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

• LLM individuation problemconcept0.828
  The central problem the paper addresses: which entities associated with LLMs, if any, should be identified as minds
• Three views are the strongest candidates for LLM individuation: the virtual instance view, the instance-persona view, and the model-persona viewclaim0.824
  Concluding thesis of the paper expanding the logical space of individuation candidates from one to three
• Linear World Models in LLMsframework0.793
  Prior work framework studying whether LLMs encode world models as linear structures in their representations
• Pluralism in LLM Persona Representationconcept0.771
  Goal of enabling models to represent and speak for diverse individuals fairly and inclusively
• LLM Personalizationconcept0.765
  Related field aiming to tailor assistant behavior to individual users, contrasted with character training's broader persona approach
• Reflection in LLMsconcept0.763
  The core phenomenon studied: the ability of LLMs to evaluate and revise their own reasoning.
• Refusal Direction in LLMsconcept0.761
  Prior finding that LLM refusal is mediated by a single latent direction, analogous to this paper's reflection direction.
• LLM Internal Representationsconcept0.760
  High-dimensional vectors produced at each transformer layer for each input token; the primary substrate analyzed in this study.