SEDGE: Structural Extrapolated Data Generation

Kun Zhang, Jiaqi Sun, Yiqing Li, Ignavier Ng, Namrata Deka, Shaoan Xie · 2026

Read SEDGE: Structural Extrapolated Data Generation by Kun Zhang, Jiaqi Sun, Yiqing Li, Ignavier Ng, Namrata Deka, Shaoan Xie (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-02482.txt.

Description

Zhang and five colleagues address generating data beyond what training data covers. That task is impossible without assumptions about how the data were produced. Their SEDGE framework states conditions under which data meeting new specs can be made reliably. It shows when the target spread can be roughly pinned down, and it gives workable methods based on structured tuning or diffusion sampling, tested on synthetic and image data.

Catalog record

CREATORKun Zhang, Jiaqi Sun, Yiqing Li, Ignavier Ng, Namrata Deka, Shaoan Xie
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE78874
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.02482

Browse the full library · About the Tower of Babel