Test Set Diversity

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

• Output Diversityconcept0.752
  The breadth of distinct outputs an LLM can produce, which is reduced by mode collapse after alignment training
• Diversity Thresholdconcept0.750
  Target minimum diversity level (e.g., 10 contradictions) that DTG iterates toward
• Multi-Response Diversityconcept0.741
  Diversity measured over a set of m responses for a single conversation, as opposed to test set diversity
• Semantic Diversity Scoremethod0.739
  Diversity metric computed as 1 minus mean pairwise cosine similarity of response embeddings, using OpenAI's text-embedding-3-small
• Diversity Parameterconcept0.739
  Gold standard value (e.g., nucleus sampling p-value) used as ground truth for evaluating diversity metrics
• Lexical Diversityconcept0.734
  Contrasted with semantic diversity; measured by n-gram overlap and distinct-n
• Semantic Diversityconcept0.730
  Core concept measured by NLI Diversity — diversity of meaning across a set of dialogue responses
• Diverse Intelligenceframework0.711
  Research program studying intelligence at multiple scales and substrates; proposed as relevant to implications of mnemonic improvisation.