Large Language Models for Patent Classification: Strengths, Trade-offs, and the Long Tail Effect

TL;DR

This study compares encoder-based models and large language models for patent classification, revealing their complementary strengths and trade-offs, especially in handling rare categories and balancing efficiency with coverage.

Contribution

It provides a systematic evaluation of LLMs versus encoder models on imbalanced patent data, highlighting the potential of hybrid approaches for improved classification.

Findings

Encoder models excel on frequent subclasses.

LLMs perform better on infrequent subclasses.

Encoder models are significantly more energy-efficient.

Abstract

Patent classification into CPC codes underpins large scale analyses of technological change but remains challenging due to its hierarchical, multi label, and highly imbalanced structure. While pre Generative AI supervised encoder based models became the de facto standard for large scale patent classification, recent advances in large language models (LLMs) raise questions about whether they can provide complementary capabilities, particularly for rare or weakly represented technological categories. In this work, we perform a systematic comparison of encoder based classifiers (BERT, SciBERT, and PatentSBERTa) and open weight LLMs on a highly imbalanced benchmark dataset (USPTO 70k). We evaluate LLMs under zero shot, few shot, and retrieval augmented prompting, and further assess parameter efficient fine tuning of the best performing model. Our results show that encoder based models achieve higher aggregate performance, driven by strong results on frequent CPC subclasses, but struggle on rare ones. In contrast, LLMs achieve relatively higher performance on infrequent subclasses, often associated with early stage, cross domain, or weakly institutionalised technologies, particularly at higher hierarchical levels. These findings indicate that encoder based and LLM based approaches play complementary roles in patent classification. We additionally quantify inference time and energy consumption, showing that encoder based models are up to three orders of magnitude more efficient than LLMs. Overall, our results inform responsible patentometrics and technology mapping, and motivate hybrid classification approaches that combine encoder efficiency with the long tail coverage of LLMs under computational and environmental constraints.