Towards Large Reasoning Models for Agriculture

2025-05-25

AI Summary

Agricultural decision-making critically depends on fine-grained contextual knowledge—spanning geography, climate, and economics—yet conventional large language models (LLMs) lack structured reasoning capabilities required for such domain-specific tasks. Method: We propose the Agricultural Large Reasoning Model (LRM) paradigm, introducing (i) AgReason—the first expert-curated, open agricultural scientific reasoning benchmark (100 questions), and (ii) AgThoughts—a large-scale dataset of 44.6K question-answer pairs with human-annotated structured reasoning chains. Leveraging foundation models (e.g., Gemini), we conduct systematic reasoning capability evaluation, supervised fine-tuning, and reasoning-chain distillation to train AgThinker, a lightweight, deployable model. Contribution/Results: The strongest Gemini baseline achieves only 36% accuracy on AgReason; in contrast, AgThinker runs efficiently on consumer-grade GPUs and significantly outperforms general-purpose LLMs. This work establishes the first comprehensive evaluation and modeling framework for agricultural reasoning, empirically validating that data-driven, domain-specific reasoning chains substantially enhance LLMs’ agricultural cognition.

Technology Category

• Knowledge Representation and Reasoning: Argumentation
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Agricultural decision-making involves complex, context-specific reasoning, where choices about crops, practices, and interventions depend heavily on geographic, climatic, and economic conditions. Traditional large language models (LLMs) often fall short in navigating this nuanced problem due to limited reasoning capacity. We hypothesize that recent advances in large reasoning models (LRMs) can better handle such structured, domain-specific inference. To investigate this, we introduce AgReason, the first expert-curated open-ended science benchmark with 100 questions for agricultural reasoning. Evaluations across thirteen open-source and proprietary models reveal that LRMs outperform conventional ones, though notable challenges persist, with the strongest Gemini-based baseline achieving 36% accuracy. We also present AgThoughts, a large-scale dataset of 44.6K question-answer pairs generated with human oversight and equipped with synthetically generated reasoning traces. Using AgThoughts, we develop AgThinker, a suite of small reasoning models that can be run on consumer-grade GPUs, and show that our dataset can be effective in unlocking agricultural reasoning abilities in LLMs. Our project page is here: https://baskargroup.github.io/Ag_reasoning/

Problem

• Enhancing agricultural decision-making with complex reasoning models
• Addressing limitations of traditional LLMs in nuanced agricultural contexts
• Developing specialized datasets and models for agricultural reasoning tasks

Innovation

• Introducing AgReason for agricultural reasoning benchmark
• Developing AgThoughts dataset with reasoning traces
• Creating AgThinker models for consumer-grade GPUs