Reproduce “EIVEN: Efficient Implicit Attribute Value Extraction using Multimodal LLM” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Henry Peng Zou, Gavin Heqing Yu, Ziwei Fan, Dan Bu, Han Liu, Peng Dai, Dongmei Jia, Cornelia Caragea
EIVEN is a multimodal framework for extracting implicit attribute values from e-commerce product data, addressing the challenge of unstated characteristics. It employs a parameter-efficient architecture and a novel 'Learning by Comparison' technique, achieving higher accuracy and greater data efficiency than previous methods while requiring fewer computational resources.
Key insights
- Attribute Value Extraction (AVE) for implicit attributes can be framed as a question-answering task, leveraging multimodal LLMs.
- Parameter-efficient fine-tuning with adapter modules can significantly reduce the computational cost of deploying multimodal LLMs for AVE.
- Learning by Comparison (LBC) techniques, which involve training on paired product comparisons, enhance a model's ability to differentiate between similar attribute values.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.