Research map / Vision-language and multimodal / Multimodal LLMs
Omni-Weather: Unified Multimodal Foundation Model for Weather Generation and Understanding
Zhiwang Zhou, Yuandong Pu, Xuming He, Yidi Liu, Yixin Chen, Junchao Gong, Xiang Zhuang, Wanghan Xu, Qinglong Cao, SHIXIANG TANG, Yihao Liu, Wenlong Zhang and 1 more
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Omni-Weather is a unified multimodal foundation model that performs both weather generation and weather understanding within a single architecture, using a radar encoder for generation tasks.
Problem
Weather modeling needs both accurate prediction and mechanistic interpretation, but existing methods treat weather generation and understanding separately.
Approach
Unifies weather generation and understanding in one multimodal foundation model by integrating a radar encoder for generation tasks alongside a multimodal backbone for understanding.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| train generation and understanding jointly rather than generation only (ablation) | Bagel-7B-MoT · SEVIR nowcasting | CSI-Mean: 0.303 -> 0.338 | Improved |
| train generation and understanding jointly rather than generation only (ablation) | Bagel-7B-MoT · SEVIR inversion | RMSE: 0.590 -> 0.514 | Improved |
| train jointly rather than understanding only (ablation) | Bagel-7B-MoT · RadarQA frame understanding | accuracy: 81.95 -> 86.65 | Improved |
| train jointly rather than understanding only (ablation) | Bagel-7B-MoT · RadarQA sequence understanding | accuracy: 54.34 -> 59.58 | Improved |
| add thinking inference to CoT-finetuned model (ablation) | Omni-Weather · SEVIR nowcasting | LPIPS: 0.182 -> 0.163 | Improved |
| add thinking inference to CoT-finetuned model (ablation) | Omni-Weather · SEVIR nowcasting | CSI-Mean: 0.347 -> 0.335 | Hurt |
| use thinking inference without CoT finetuning (ablation) | Omni-Weather · SEVIR nowcasting | CSI-Mean: 0.347 -> 0.237 | Hurt |
| add general MetaQuery data to SEVIR finetuning (ablation) | Bagel-7B-MoT · SEVIR nowcasting | CSI-Mean: 0.3352 -> 0.3471 | Improved |
| add general MetaQuery data to SEVIR finetuning (ablation) | Bagel-7B-MoT · SEVIR nowcasting | CRPS: 0.0231 -> 0.0226 | Improved |
| unified Omni-Weather vs specialized nowcasting baselines | Omni-Weather · SEVIR nowcasting | CRPS: 0.031 -> 0.026 | Improved |
| unified model vs RadarQA on image understanding miss accuracy | Omni-Weather · RadarQA image understanding | Miss accuracy: 67.67 -> 92.21 | Improved |
| unified model vs RadarQA on sequence understanding | Omni-Weather · RadarQA sequence understanding | Overall accuracy: 66.17 -> 61.79 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- RadarQA — RadarQA: Multi-modal Quality Analysis of Weather Radar Forecasts (NeurIPS 2025) · better on 3 of 4 shared results
- WeatherGFM — WeatherGFM: Learning a Weather Generalist Foundation Model via In-context Learning (ICLR 2025) · better on 1 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- WeatherGFM: Learning a Weather Generalist Foundation Model via In-context Learning (ICLR 2025 · similarity 0.88)
- OmniCast: A Masked Latent Diffusion Model for Weather Forecasting Across Time Scales (NeurIPS 2025 · similarity 0.85)
- FutureOmni: Evaluating Future Forecasting from Omni-Modal Context for Multimodal LLMs (ICML 2026 · similarity 0.84)
- WeatherPrompt: Multi-modality Representation Learning for All-Weather Drone Visual Geo-Localization (NeurIPS 2025 · similarity 0.84)
- WeatherSyn: An Instruction Tuning MLLM For Weather Forecasting Report Generation (ICML 2026 · similarity 0.84)
- ThinkOmni: Lifting Textual Reasoning to Omni-modal Scenarios via Guidance Decoding (ICLR 2026 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
