LLM-Based Facet Annotation

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

• LLM Personalizationconcept0.768
  Related field aiming to tailor assistant behavior to individual users, contrasted with character training's broader persona approach
• LLM-Judge Data Attributionmethod0.725
  Alternative data attribution approach using an LLM as a judge; compared against the probe-based method.
• Large Language Models (LLMs)concept0.725
  Transformer-based models like GPT-4, LaMDA, PaLM; assessed for GWT indicators.
• Feature completeness search using LLM-generated queriesmethod0.724
  Using Claude to search for features activating on specific concepts and automated labeling.
• Linear World Models in LLMsframework0.718
  Prior work framework studying whether LLMs encode world models as linear structures in their representations
• Linear Representation of Concepts in LLMsconcept0.717
  The finding that interpretable concepts including character traits are encoded as linear directions in transformer residual streams
• do facet-level control vectors learned from one backbone transfer to other LLM architectures?question0.716
  Open question about transferability implicit in the two-backbone evaluation design
• Automated interpretability using LLMs can usefully score feature specificity.claim0.714
  Claude 3 Opus ratings aligned with human judgment of feature descriptions.