PatentVision: A multimodal method for drafting patent applications

2025-10-10

AI Summary

Patent drafting faces three major challenges: technical complexity, stringent legal compliance requirements, and difficulty in cross-modal alignment between textual descriptions and figures. This paper introduces large vision-language models (LVLMs) to automated patent specification generation for the first time, proposing a domain-specific multimodal generation framework. It employs domain-adaptive fine-tuning to jointly model claims text and accompanying drawings by integrating fine-grained visual understanding with patent-domain knowledge. Experiments demonstrate that our approach significantly outperforms text-only baselines in claim accuracy, completeness of design feature representation, and fidelity of function–structure correspondence—yielding outputs approaching human-drafted quality. Our core contribution is the establishment of the first LVLM paradigm for patent generation that explicitly supports text–figure co-generation, empirically validated for effectiveness and practical utility in real-world patent drafting scenarios.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Large Vision Models
• Natural Language Processing: Generation

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Patent drafting is complex due to its need for detailed technical descriptions, legal compliance, and visual elements. Although Large Vision Language Models (LVLMs) show promise across various tasks, their application in automating patent writing remains underexplored. In this paper, we present PatentVision, a multimodal framework that integrates textual and visual inputs such as patent claims and drawings to generate complete patent specifications. Built on advanced LVLMs, PatentVision enhances accuracy by combining fine tuned vision language models with domain specific training tailored to patents. Experiments reveal it surpasses text only methods, producing outputs with greater fidelity and alignment with human written standards. Its incorporation of visual data allows it to better represent intricate design features and functional connections, leading to richer and more precise results. This study underscores the value of multimodal techniques in patent automation, providing a scalable tool to reduce manual workloads and improve consistency. PatentVision not only advances patent drafting but also lays the groundwork for broader use of LVLMs in specialized areas, potentially transforming intellectual property management and innovation processes.

Problem

• Automating patent drafting with multimodal text and visual inputs
• Enhancing accuracy through domain-specific vision-language model training
• Generating patent specifications with improved fidelity and legal compliance

Innovation

• Multimodal framework integrates textual and visual inputs
• Fine tuned vision language models enhance patent accuracy
• Generates complete specifications surpassing text only methods