Text-to-Distribution Prediction with Quantile Tokens and Neighbor Context

Yilun Zhu, Yuan Zhuang, Nikhita Vedula, Dushyanta Dhyani, Shaoyuan Xu, Moyan Li, Mohsen Bayati, Bryan Wang, Shervin Malmasi · 2026

Read Text-to-Distribution Prediction with Quantile Tokens and Neighbor Context by Yilun Zhu, Yuan Zhuang, Nikhita Vedula, Dushyanta Dhyani, Shaoyuan Xu, Moyan Li, Mohsen Bayati, Bryan Wang, Shervin Malmasi (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-2604-20216.txt.

Description

Zhu and eight colleagues introduce Quantile Token Regression, predicting full outcome spreads from text, not single values. Each input pairs with several seen quantile outcomes, and the target is a dense grid of quantiles. Special quantile tokens in the input let each quantile link straight to the input, fixing weak local grounding and shared bottlenecks. Retrieved neighbor cases with their own spreads sharpen estimates further. Tests on Inside Airbnb and StackSample show about four points lower error and twice as narrow bands, with top gains on small hard sets.

Catalog record

CREATORYilun Zhu, Yuan Zhuang, Nikhita Vedula, Dushyanta Dhyani, Shaoyuan Xu, Moyan Li, Mohsen Bayati, Bryan Wang, Shervin Malmasi
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE61134
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.20216

Browse the full library · About the Tower of Babel