18 Nov 2025
Alimama's MOON framework enhances e-commerce search advertising by developing high-quality multimodal representations through a decoupled, three-stage training paradigm utilizing a domain-adapted Multimodal Large Language Model. Deployed across Taobao, it delivered a +20.00% overall online Click-Through Rate improvement, with specific uplifts of +34.80% for new products and +35.74% in the fashion category.
04 Dec 2025
Alibaba Group researchers developed LORE, a large generative model framework for e-commerce search relevance, which systematically deconstructs the relevance task and employs a two-stage training paradigm. The framework successfully integrates knowledge, multi-modal understanding, and rule adherence, leading to a cumulative +27% improvement in the online GoodRate metric and outperforming state-of-the-art LLMs on a custom e-commerce benchmark.
There are no more papers matching your filters at the moment.
