LLMs Are Also Effective Embedding Models: An In-depth Overview

By Chong-Yang Tao, Tao Shen, Shen Gao, Jun-Shuo Zhang, Zhen Li, Kai Hua, Wen-Peng Hu, Zhengwei Tao, Shuai Ma. 2024

Subfields: Large Language Models, Representation & Self-Supervised Learning

Citations in the corpus

• Embedding Models Measure in Peculiar Ways (2026, unknown): “While physical measurements have been explored in LLMs (Park et al., 2022) and word embeddings (Sundararaman et al., 2020), our exploration is conducted on the contrastively trained text representations that promise to preserve “essential semantic and syntactic information” (Tao et al., 2026).”
• Understanding generative AI output with embedding models (2025, unknown): “The FE produces embedding vectors from raw input data (text, images, etc.) and typically encompasses all but the final layer of the DNN ( 17 , 18 ).”
• LLM-Enhanced Semantic Text Segmentation (2025, unknown): “Although a wide range of LLM-based models is available [15,16], in this study, we restrict attention to models that have been specifically trained or optimized for embedding tasks rather than text generation.”
• Unlocking Multimodal Document Intelligence: From Current Triumphs to Future Frontiers of Visual Document Retrieval (2026, unknown): “This includes the development of MLLM-based embedding and reranker models to enhance semantic matching (Tao et al., 2024; Zhang et al., 2024b; Wang et al., 2025d).”
• CoRet: Improved Retriever for Code Editing (2025, unknown): “Several modern LLMs have been modified to output feature embeddings (BehnamGhader et al., 2024; Tao et al., 2024).”