Diversity Tuning via Probability Threshold

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

• Diversity Threshold Generationmethod0.824
  Iterative generation procedure that resamples lowest-scoring responses until a diversity threshold is reached
• Diversity Thresholdconcept0.811
  Target minimum diversity level (e.g., 10 contradictions) that DTG iterates toward
• Fine-Tuning Threshold Recalibrationmethod0.781
  Re-running probabilistic bisection on each fine-tuned checkpoint to normalize first-attempt difficulty
• Diversity Threshold Generation increases semantic diversity with minimal loss in relevancyclaim0.769
  Central practical claim about the DTG procedure
• Diversity Threshold Generation results in average 137% increase in NLI Diversity compared to standard generationfinding0.761
  Key headline result of the DTG procedure across all conditions
• Hypothesis 1 (Threshold Behavior): There exists a task-dependent threshold Sc such that performance exhibits sharp changes as S crosses Sc, with value and transition width depending on model, layer, and poolinghypothesis0.744
  Core testable hypothesis of UCCT about the nature of performance transitions under anchoring
• Typicality bias rates in instruction-tuned models remain at similar or higher levels compared to their base model counterpartsfinding0.743
  Shows typicality bias is preserved through instruction tuning and RLHF, not introduced by alignment
• The between-to-within-class variance ratio peaks at different layers for different tasks, confirming no single layer is universally optimal.finding0.737
  Supports the claim against single-layer probing approaches used in prior work.