← 研踪 PaperTrailarXiv:2611.25570 · v1
Reinforcement-Aligned Probabilistic Weather Forecasting with Discrete Tokens
首次提交 2026-11-19 14:10
版本更新 2026-11-19 14:10
雷达观测 2026-11-20 19:58
摘要
Global medium-range weather forecasting must not only generate ensembles efficiently to characterize uncertainty in future states, but also serve downstream applications such as extreme-event warnings and cross-sphere coupled simulations. We introduce WeatherCraft, a framework that unifies probabilistic generation and reinforcement learning alignment through discrete weather tokens, bringing the GPT-style pretraining--post-training paradigm to global medium-range weather forecasting. A multiscale tokenizer represents atmospheric states as a coarse-to-fine token hierarchy, while a Transformer models causal dependencies across time and scale and predicts tokens within each scale in parallel. Direct sampling generates diverse ensembles, and exact token log-probabilities enable Group Relative Policy Optimization (GRPO) to maximize downstream task rewards without differentiating through evaluators or coupled models. On the ERA5 benchmark, WeatherCraft's eight-member ensemble mean reduces day-15 normalized RMSE by 17.6% relative to the strongest evaluated deterministic baseline and, with KV caching, achieves an approximately inference speedup over GenCast. Direct fair-CRPS adaptation improves ensemble probabilistic skill, while reward-based post-training improves forecasts of persistent land temperature extremes and coupled ocean heat extremes. This work establishes discrete autoregressive modeling as a general foundation for scalable, task-alignable Earth system foundation models.
作者备注:28 pages, 11 figures, 15 tables. Code: https://github.com/Jinhan-Li/WeatherCraft
观测到的版本
相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。
v1 · 2026-11-19 14:10 · 最新观测版本
Reinforcement-Aligned Probabilistic Weather Forecasting with Discrete Tokens
Global medium-range weather forecasting must not only generate ensembles efficiently to characterize uncertainty in future states, but also serve downstream applications such as extreme-event warnings and cross-sphere coupled simulations. We introduce WeatherCraft, a framework that unifies probabilistic generation and reinforcement learning alignment through discrete weather tokens, bringing the GPT-style pretraining--post-training paradigm to global medium-range weather forecasting. A multiscale tokenizer represents atmospheric states as a coarse-to-fine token hierarchy, while a Transformer models causal dependencies across time and scale and predicts tokens within each scale in parallel. Direct sampling generates diverse ensembles, and exact token log-probabilities enable Group Relative Policy Optimization (GRPO) to maximize downstream task rewards without differentiating through evaluators or coupled models. On the ERA5 benchmark, WeatherCraft's eight-member ensemble mean reduces day-15 normalized RMSE by 17.6% relative to the strongest evaluated deterministic baseline and, with KV caching, achieves an approximately inference speedup over GenCast. Direct fair-CRPS adaptation improves ensemble probabilistic skill, while reward-based post-training improves forecasts of persistent land temperature extremes and coupled ocean heat extremes. This work establishes discrete autoregressive modeling as a general foundation for scalable, task-alignable Earth system foundation models.
查看此版本解读 →查看此版本原文 ↗