Weather-R1: Logically Consistent Reinforcement Fine-Tuning for Multimodal Reasoning in Meteorology

TL;DR

Weather-R1 introduces a logically consistent multimodal reasoning model for meteorology, utilizing a new benchmark and a novel reinforcement fine-tuning method to improve reasoning faithfulness and performance.

Contribution

The paper presents Weather-R1, the first reasoning VLM with logical faithfulness in meteorology, and proposes LoCo-RFT, a reinforcement fine-tuning approach with a logical consistency reward.

Findings

Weather-R1 improves WeatherQA performance by 9.8 percentage points.

LoCo-RFT effectively resolves reasoning contradictions in the model.

Weather-R1 outperforms baseline models including Supervised Fine-Tuning and RFT.

Abstract

While Vision Language Models (VLMs) show advancing reasoning capabilities, their application in meteorology is constrained by a domain gap and a reasoning faithfulness gap. Specifically, mainstream Reinforcement Fine-Tuning (RFT) can induce Self-Contradictory Reasoning (Self-Contra), where the model's reasoning contradicts its final answer, which is unacceptable in such a high-stakes domain. To address these challenges, we construct WeatherQA, a novel multimodal reasoning benchmark in meteorology. We also propose Logically Consistent Reinforcement Fine-Tuning (LoCo-RFT), which resolves Self-Contra by introducing a logical consistency reward. Furthermore, we introduce Weather-R1, the first reasoning VLM with logical faithfulness in meteorology, to the best of our knowledge. Experiments demonstrate that Weather-R1 improves performance on WeatherQA by 9.8 percentage points over the baseline, outperforming Supervised Fine-Tuning and RFT, and even surpassing the original Qwen2.5-VL-32B. These results highlight the effectiveness of our LoCo-RFT and the superiority of Weather-R1. Our benchmark and code are available at https://github.com/Marcowky/Weather-R1.