Saved in:
Bibliographic Details
Main Authors: Choi, Sunguk, Kwon, Yonghoon, Lee, Heondeuk
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2508.18743
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908539248705536
author Choi, Sunguk
Kwon, Yonghoon
Lee, Heondeuk
author_facet Choi, Sunguk
Kwon, Yonghoon
Lee, Heondeuk
contents Long chain-of-thought (CoT) prompting helps Large Language Models (LLMs) solve difficult problems, but very long traces often slow or even degrade performance on fast, intuitive "System-1" tasks. We introduce Connector-Aware Compact CoT (CAC-CoT) -- a method that deliberately restricts reasoning to a small, fixed set of connector phrases, steering the model toward concise and well -- structured explanations. Despite its simplicity, our synthetic method with general-purpose LLMs yields a high-quality training quality. CAC-CoT achieves approximately 85% on GSM8K and approximately 40% on GPQA (System-2) while also achieving approximately 85% on S1-Bench (System-1), surpassing the baseline by over 20%. Its reasoning traces average approximately 300 tokens(ART), about one-third the length of baseline traces, delivering higher efficiency without loss of accuracy.
format Preprint
id arxiv_https___arxiv_org_abs_2508_18743
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle CAC-CoT: Connector-Aware Compact Chain-of-Thought for Efficient Reasoning Data Synthesis Across Dual-System Cognitive Tasks
Choi, Sunguk
Kwon, Yonghoon
Lee, Heondeuk
Artificial Intelligence
Computation and Language
Long chain-of-thought (CoT) prompting helps Large Language Models (LLMs) solve difficult problems, but very long traces often slow or even degrade performance on fast, intuitive "System-1" tasks. We introduce Connector-Aware Compact CoT (CAC-CoT) -- a method that deliberately restricts reasoning to a small, fixed set of connector phrases, steering the model toward concise and well -- structured explanations. Despite its simplicity, our synthetic method with general-purpose LLMs yields a high-quality training quality. CAC-CoT achieves approximately 85% on GSM8K and approximately 40% on GPQA (System-2) while also achieving approximately 85% on S1-Bench (System-1), surpassing the baseline by over 20%. Its reasoning traces average approximately 300 tokens(ART), about one-third the length of baseline traces, delivering higher efficiency without loss of accuracy.
title CAC-CoT: Connector-Aware Compact Chain-of-Thought for Efficient Reasoning Data Synthesis Across Dual-System Cognitive Tasks
topic Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2508.18743