Fast Quiet-STaR: Thinking Without Thought Tokens

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Huang, Wei, Xiong, Yizhe, Ye, Xin, Deng, Zhijie, Chen, Hui, Lin, Zijia, Ding, Guiguang
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866908521839198208
author Huang, Wei
Xiong, Yizhe
Ye, Xin
Deng, Zhijie
Chen, Hui
Lin, Zijia
Ding, Guiguang
author_facet Huang, Wei
Xiong, Yizhe
Ye, Xin
Deng, Zhijie
Chen, Hui
Lin, Zijia
Ding, Guiguang
contents Large Language Models (LLMs) have achieved impressive performance across a range of natural language processing tasks. However, recent advances demonstrate that further gains particularly in complex reasoning tasks require more than merely scaling up model sizes or training data. One promising direction is to enable models to think during the reasoning process. Recently, Quiet STaR significantly improves reasoning by generating token-level thought traces, but incurs substantial inference overhead. In this work, we propose Fast Quiet STaR, a more efficient reasoning framework that preserves the benefits of token-level reasoning while reducing computational cost. Our method introduces a curriculum learning based training strategy that gradually reduces the number of thought tokens, enabling the model to internalize more abstract and concise reasoning processes. We further extend this approach to the standard Next Token Prediction (NTP) setting through reinforcement learning-based fine-tuning, resulting in Fast Quiet-STaR NTP, which eliminates the need for explicit thought token generation during inference. Experiments on four benchmark datasets with Mistral 7B and Qwen2.5 7B demonstrate that Fast Quiet-STaR consistently outperforms Quiet-STaR in terms of average accuracy under the same inference time budget. Notably, Fast Quiet-STaR NTP achieves an average accuracy improvement of 9\% on Mistral 7B and 5.7\% on Qwen2.5 7B, while maintaining the same inference latency. Our code will be available at https://github.com/huangwei200012/Fast-Quiet-STaR.
format Preprint
id arxiv_https___arxiv_org_abs_2505_17746
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Fast Quiet-STaR: Thinking Without Thought Tokens
Huang, Wei
Xiong, Yizhe
Ye, Xin
Deng, Zhijie
Chen, Hui
Lin, Zijia
Ding, Guiguang
Computation and Language
68T50
I.2.7
Large Language Models (LLMs) have achieved impressive performance across a range of natural language processing tasks. However, recent advances demonstrate that further gains particularly in complex reasoning tasks require more than merely scaling up model sizes or training data. One promising direction is to enable models to think during the reasoning process. Recently, Quiet STaR significantly improves reasoning by generating token-level thought traces, but incurs substantial inference overhead. In this work, we propose Fast Quiet STaR, a more efficient reasoning framework that preserves the benefits of token-level reasoning while reducing computational cost. Our method introduces a curriculum learning based training strategy that gradually reduces the number of thought tokens, enabling the model to internalize more abstract and concise reasoning processes. We further extend this approach to the standard Next Token Prediction (NTP) setting through reinforcement learning-based fine-tuning, resulting in Fast Quiet-STaR NTP, which eliminates the need for explicit thought token generation during inference. Experiments on four benchmark datasets with Mistral 7B and Qwen2.5 7B demonstrate that Fast Quiet-STaR consistently outperforms Quiet-STaR in terms of average accuracy under the same inference time budget. Notably, Fast Quiet-STaR NTP achieves an average accuracy improvement of 9\% on Mistral 7B and 5.7\% on Qwen2.5 7B, while maintaining the same inference latency. Our code will be available at https://github.com/huangwei200012/Fast-Quiet-STaR.
title Fast Quiet-STaR: Thinking Without Thought Tokens
topic Computation and Language
68T50
I.2.7
url https://arxiv.org/abs/2505.17746