Table 3 Model selection

From: Visual information identification and Q&A of intangible cultural heritage inheritors by using enhanced Graph-Retrieval framework

Model category

Model

Model introduction

LLMs

GPT-4

GPT-4 is the latest generation of generative pre-trained models released by OpenAI, which demonstrates outstanding performance in language modeling and generation tasks.

Llama-3

(Llama-3-7b-chat)

Llama-3 is a large-scale language model released by Meta, known for its efficient model structure and training methods.

ChatGLM-4

ChatGLM-4 is a generative dialogue model proposed by Zhipu AI company, which performs excellently in dialogue generation and language understanding.

RAG-LLm

RAG-Llama3

RAG-GPT4

RAG-ChatGLM4

Introducing the RAG strategy constructed in this article, which combines retrieval knowledge, contextual information, and prompt examples to guide LLMs to better adapt to downstream tasks.

Back to article page