Question

Description acceptable. Over 100 implementations of a strategy for this task are surveyed in a 2024 paper by Gao et al. that categorizes them as Naive, Advanced, or Modular. Novel embeddings of query-specific documents are indexed and used in the “retrieval-augmented” form of this task. This task is performed (10[1])by applying Brownian noise to points in a latent space (10[1])and then decoding in diffusion models. A game-theoretic (10[1])agent improves at this task while competing against a discriminator in a type of “adversarial” neural (10[1])network. (10[1])This task names a framework consisting of unsupervised “pre-training” followed by supervised “fine-tuning.” Nonsensical outputs during this task are called “hallucinations.” (10[1])For 10 points, what task names a form of AI commonly used to produce text and images (10[1])that is the “G” (10[1])in GPT? ■END■

ANSWER: generation [or word forms such as generate; accept text generation or image generation; accept generative artificial intelligence or generative AI or generative pre-trained transformer or generative adversarial network or retrieval-augmented generation; accept descriptions such as computer-produced text or equivalents]
<Johns Hopkins, Other Science>
= Average correct buzz position

Back to tossups

Buzzes

PlayerTeamOpponentBuzz PositionValue
Vinayak Singh BhadoriyaNYU BRutgers A4810
Jerry VinokurovJohn Jay CollegeMaryland B5810
Derek ChenColumbia CVassar A6610
Ryan RosenbergNYU AGeorge Washington B8210
Danny HanPenn AHaverford B8310
Isaac MammelMaryland APenn B10410
Avery BarnettHaverford AYale A12110
Nathan ZhangCornell BGeorge Washington A12510

Summary

California2025-02-01Y3100%0%33%99.00
Lower Mid-Atlantic2025-02-01Y6100%0%33%114.33
Midwest2025-02-01Y6100%0%33%93.67
Northeast2025-02-01Y5100%0%20%95.60
Pacific Northwest2025-02-01Y2100%0%50%116.00
South Central2025-02-01Y2100%0%0%44.00
Southeast2025-02-01Y1100%0%100%128.00
Upper Mid-Atlantic2025-02-01Y8100%0%0%85.88