Refine Thought: A Test-Time Inference Method for Embedding Model Reasoning

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Wang, Guangzhi, Li, Kai, Jiao, Yinghao, Liu, Zhi
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866911271187644416
author Wang, Guangzhi
Li, Kai
Jiao, Yinghao
Liu, Zhi
author_facet Wang, Guangzhi
Li, Kai
Jiao, Yinghao
Liu, Zhi
contents We propose RT (Refine Thought), a method that can enhance the semantic rea-soning ability of text embedding models. The method obtains the final semanticrepresentation by running multiple forward passes of the text embedding model.Experiments show that RT achieves significant improvements on semantic reason-ing tasks in BRIGHT and the person job matching benchmark PJBenchmark1, while maintaining consistent performance on general-purpose semantic under-standing tasks such as C-MTEB. Our results indicate that RT is effective becauseit further activates the semantic reasoning ability learned during pretraining bydecoder-only text embedding models(e.g., Qwen3-Embedding-8B). RT canbe seen as a test-time inference method.
format Preprint
id arxiv_https___arxiv_org_abs_2511_13726
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Refine Thought: A Test-Time Inference Method for Embedding Model Reasoning
Wang, Guangzhi
Li, Kai
Jiao, Yinghao
Liu, Zhi
Computation and Language
Artificial Intelligence
Information Retrieval
We propose RT (Refine Thought), a method that can enhance the semantic rea-soning ability of text embedding models. The method obtains the final semanticrepresentation by running multiple forward passes of the text embedding model.Experiments show that RT achieves significant improvements on semantic reason-ing tasks in BRIGHT and the person job matching benchmark PJBenchmark1, while maintaining consistent performance on general-purpose semantic under-standing tasks such as C-MTEB. Our results indicate that RT is effective becauseit further activates the semantic reasoning ability learned during pretraining bydecoder-only text embedding models(e.g., Qwen3-Embedding-8B). RT canbe seen as a test-time inference method.
title Refine Thought: A Test-Time Inference Method for Embedding Model Reasoning
topic Computation and Language
Artificial Intelligence
Information Retrieval
url https://arxiv.org/abs/2511.13726