What type of study is this?

This is a Quantitative Study study.

September 28, 2025Open Access

A Scalable Multi-Agent Framework for Low-Resource E-Commerce Concept Extraction and Standardization

Key Points

The multi-agent framework significantly enhances understanding of shopping concepts in e-commerce, improving clarity in consumer behavior.
Using LLaMA-2 70B, the framework integrates retrieval-augmented generation and cross-modal feature fusion for effective task handling.
Domain-adaptive pretraining and multi-step training methods adapt well to low-resource situations, achieving balanced results.
Application of the framework in real e-commerce demonstrates its promising potential for practical deployment despite data limitations.

Abstract

Understanding shopping concepts in e-commerce is hard because consumer behavior, product terms, and technical details vary a lot. This paper presents MALLM, a new multi-agent framework based on LLaMA-2 70B. It improves concept understanding by letting specialized agents work together. The model uses a layered agent system, domain-adaptive pretraining, retrieval-augmented generation (RAG), and cross-modal feature fusion to handle tasks clearly and well. It also applies knowledge distillation and multi-step training to adapt to small data. MALLM balances good results with practical deployment. It works well in real e-commerce situations.

Read Full Paperexternally

Bookmark

View Full Paper

Cite This Study

Aijia Sun (Fri,) studied this question.

synapsesocial.com/papers/68d9052541e1c178a14f5689 https://doi.org/https://doi.org/10.20944/preprints202509.2108.v1