VS increases diversity by 1.6-2.1x over direct prompting on creative writing tasks (poem, story, joke)

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

• VS improves human evaluation scores by 25.7% on creative writing compared to direct promptingfinding0.882
  Human study result validating automatic diversity metrics for creative writing tasks
• VS-Standard achieves human-rated diversity scores of 2.39/3.06/3.01 for poem/story/joke vs. 1.90/2.74/1.83 for Direct prompting on 4-point scalefinding0.849
  Human study confirming automatic diversity metrics align with human perceptions
• VS-Standard achieves 1.86x diversity gain at only 1.12x cost increase over Direct prompting on poem generationfinding0.849
  Cost-diversity trade-off analysis showing VS's practical efficiency
• After DPO stage, VS outperforms direct prompting by 182.6% on diversity in poem continuation using Tulu-70Bfinding0.830
  Demonstrates the magnitude of VS's advantage over direct prompting after aggressive alignment training
• VS-CoT and VS-Multi push the Pareto front of the diversity-quality trade-off across creative writing tasksclaim0.804
  Empirical finding from creative writing experiments showing VS variants achieve higher diversity without sacrificing quality
• VS recovers 66.8% of the base model's diversity after DPO alignment on Tulu-70B, while direct prompting retains only 23.8%finding0.784
  Quantifies how much of the base model's diversity VS can recover compared to baseline prompting
• In evaluating creative or intellectual output, the question should be 'Does it elevate us?' rather than 'What made it?' — judging origin rather than quality emphasizes the worst parts of human nature.claim0.770
  Normative claim about how to evaluate AI-generated content, using Deutsche Physik as cautionary analogy
• All VS variants maintain refusal rates above 97% on StrongReject benchmark, within 0.3-0.8 percentage points of the Direct baselinefinding0.757
  Confirms VS does not compromise safety alignment while improving diversity