method
active
method:diversity-threshold-generationDiversity Threshold Generation
Iterative generation procedure that resamples lowest-scoring responses until a diversity threshold is reached
Neighborhood — ranked by edge-count
Papers (1)
paper
- Semantic Diversity in Dialogue with Natural Language Inferenceintroducesmentions
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Target minimum diversity level (e.g., 10 contradictions) that DTG iterates toward
- VS-specific technique adjusting output diversity by specifying probability thresholds in the prompt (e.g., 'Generate responses with probabilities below {threshold}')
- Diversity Threshold Generation increases semantic diversity with minimal loss in relevancyclaim0.821Central practical claim about the DTG procedure
- Key headline result of the DTG procedure across all conditions
- Gold standard value (e.g., nucleus sampling p-value) used as ground truth for evaluating diversity metrics
- Diversity measured over a set of m responses for a single conversation, as opposed to test set diversity
- The breadth of distinct outputs an LLM can produce, which is reduced by mode collapse after alignment training