finding
active
finding:average-percent-difference-in-bertscore-from-starting-to-ending-dtg-sets-is-0-08Average percent difference in BERTScore from starting to ending DTG sets is 0.08%
Shows relevancy is maintained by Diversity Threshold Generation
Source paper
extracted_from(2022) · Katherine Stasaski · Marti A. Hearst
Neighborhood — ranked by edge-count
Papers (1)
paper
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.
- High cosine similarity for Gemma3 steering vectors suggests strong linear reflection structure.
- Confirms nucleus sampling produces more semantically diverse outputs than beam search
- Shows VS substantially better approximates the pretraining distribution than baseline methods
- Average faithfulness 3.6x higher for Claude series and 1.7x for DeepSeek series (reasoning vs non-reasoning)finding0.703Quantifies reasoning advantage.
- At S=200 variation across runs (TVD 0.0451) nearly equals sampling variation within a run (TVD 0.0442).finding0.701Supports multinomial model.
- 20,000 trials.
- Dataset mixture experiments establish the fraction of incorrect data needed to induce misalignment
- Base corrections: 72.9% Failed Fix, 14.7% Fixed; retractions: 69.4% Correct Retract, 30.0% Incorrect Retractfinding0.699Label distributions.