【Proposal】ADSFT: Adaptive and Domain-Specific Fine-Tuning Paradigms for Generative Models
Keywords: LM, Finetune
TL;DR: the goal of this research is to fine-tune ChatGLM3-6B to generate e-commerce advertisements more quickly and accurately, improving the overall quality and efficiency of the advertisement generation process.
Abstract: In the e-commerce industry, generating high-quality advertisements tailored to specific products is
crucial for enhancing product visibility and improving sales conversion rates. The Amazon Product
dataset offers a rich set of product information, including keywords and key-value attribute lists
(kv-list), that supports the creation of precise advertisement generation tasks. However, the generated
advertisements need to ensure both high generation speed and factual consistency with the input
product information. Therefore, the goal of this research is to fine-tune ChatGLM3-6B to generate
e-commerce advertisements more quickly and accurately, improving the overall quality and efficiency
of the advertisement generation process.
Submission Number: 56
Loading