Generative Multimodal Entity Linking
This work addresses the scalability and cost issues in MEL for researchers and practitioners by providing an efficient solution that leverages LLMs with minimal fine-tuning.
The paper tackles the problem of Multimodal Entity Linking (MEL) by proposing GEMEL, a generative framework based on Large Language Models (LLMs) that directly generates entity names, achieving state-of-the-art results with 7.7% and 8.8% accuracy gains on two datasets while fine-tuning only about 0.3% of model parameters.
Multimodal Entity Linking (MEL) is the task of mapping mentions with multimodal contexts to the referent entities from a knowledge base. Existing MEL methods mainly focus on designing complex multimodal interaction mechanisms and require fine-tuning all model parameters, which can be prohibitively costly and difficult to scale in the era of Large Language Models (LLMs). In this work, we propose GEMEL, a Generative Multimodal Entity Linking framework based on LLMs, which directly generates target entity names. We keep the vision and language model frozen and only train a feature mapper to enable cross-modality interactions. To adapt LLMs to the MEL task, we leverage the in-context learning capability of LLMs by retrieving multimodal instances as demonstrations. Extensive experiments show that, with only ~0.3% of the model parameters fine-tuned, GEMEL achieves state-of-the-art results on two well-established MEL datasets (7.7% accuracy gains on WikiDiverse and 8.8% accuracy gains on WikiMEL). The performance gain stems from mitigating the popularity bias of LLM predictions and disambiguating less common entities effectively. Further analysis verifies the generality and scalability of GEMEL. Our framework is compatible with any off-the-shelf language model, paving the way towards an efficient and general solution for utilizing LLMs in the MEL task. Our code is available at https://github.com/HITsz-TMG/GEMEL.