SynCast: Synergizing Contradictions in Precipitation Nowcasting via Diffusion Sequential Preference Optimization
Kaiyi Xu, Junchao Gong, Wenlong Zhang, Ben Fei, Lei Bai, Wanli Ouyang · 2025
Read SynCast: Synergizing Contradictions in Precipitation Nowcasting via Diffusion Sequential Preference Optimization by Kaiyi Xu, Junchao Gong, Wenlong Zhang, Ben Fei, Lei Bai, Wanli Ouyang (2025) 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-2510-21847.txt.
Description
Kaiyi Xu, Junchao Gong, Wenlong Zhang, Ben Fei, Lei Bai, and Wanli Ouyang address a tug-of-war in short-term rain forecasting from radar. Smooth, cautious forecasts miss extreme downpours, while bold generative forecasts wobble between scoring measures, and two standard measures reward opposite habits: catching events versus not crying wolf.
Their system, SynCast, borrows preference training from language modeling and applies it to a diffusion forecaster in two stages. Training first teaches the model to suppress false alarms, then improves event detection while holding the false-alarm gains in place. The staged approach reconciles measures that usually trade off against each other, giving steadier performance across both. The work is the first to bring preference optimization into precipitation nowcasting.
Catalog record
| CREATOR | Kaiyi Xu, Junchao Gong, Wenlong Zhang, Ben Fei, Lei Bai, Wanli Ouyang |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 51 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2510-21847