Agri-R1: Empowering Generalizable Agricultural Reasoning in Vision-Language Models with Reinforcement Learning

2026-01-08

arXiv.org

AI Summary

This study addresses the limitations of existing vision-language models in agricultural disease diagnosis—namely, their reliance on strong annotations, poor interpretability, and weak generalization in open-ended scenarios. The authors propose a novel method that automatically generates reasoning data without manual labeling by integrating vision-language synthesis with large language model filtering, constructing a high-quality training set using only 19% of the original samples. They further introduce a new reward function combining domain-specific lexicons and fuzzy matching, enabling structured reasoning through Group Relative Policy Optimization (GRPO). Evaluated on CDDMBench, their 3B-parameter model substantially outperforms 7B–13B baselines, achieving a 23.2% gain in disease identification accuracy, a 33.3% improvement in agricultural question answering, and a 26.10-point increase in cross-domain generalization.

Technology Category

• Computer Vision: Language and Vision
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Agricultural disease diagnosis challenges VLMs, as conventional fine-tuning requires extensive labels, lacks interpretability, and generalizes poorly. While reasoning improves model robustness, existing methods rely on costly expert annotations and rarely address the open-ended, diverse nature of agricultural queries. To address these limitations, we propose \textbf{Agri-R1}, a reasoning-enhanced large model for agriculture. Our framework automates high-quality reasoning data generation via vision-language synthesis and LLM-based filtering, using only 19\% of available samples. Training employs Group Relative Policy Optimization (GRPO) with a novel proposed reward function that integrates domain-specific lexicons and fuzzy matching to assess both correctness and linguistic flexibility in open-ended responses. Evaluated on CDDMBench, our resulting 3B-parameter model achieves performance competitive with 7B- to 13B-parameter baselines, showing a +23.2\% relative gain in disease recognition accuracy, +33.3\% in agricultural knowledge QA, and a +26.10-point improvement in cross-domain generalization over standard fine-tuning. Ablation studies confirm that the synergy between structured reasoning data and GRPO-driven exploration underpins these gains, with benefits scaling as question complexity increases.

Problem

• agricultural disease diagnosis
• vision-language models
• open-ended reasoning
• data annotation
• cross-domain generalization

Innovation

• reinforcement learning
• vision-language models
• reasoning data synthesis
• Group Relative Policy Optimization
• agricultural disease diagnosis