发表机构
Institute for Logic, Language and Computation, University of Amsterdam; INDE Lab, University of Amsterdam（阿姆斯特丹大学逻辑、语言与计算研究所; 阿姆斯特丹大学INDE实验室）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本研究针对思维模式融合（TMF）训练动态开展系统性研究，揭示两种模式间的非对称相互作用与负相关关系，为设计TMF训练设置提供指导。
AI 中文摘要
思维模式融合（TMF）使大型语言模型通过在单一模型内统一非思维模式与思维模式，同时支持简洁响应与长形式推理。然而，其训练动态（包括两种模式间的数据比例和训练计划）仍未得到充分探索。本研究通过分析思维模式与非思维模式间的训练计划和数据比例，对TMF开展系统性研究。聚焦数学问题解决，我们构建了包含多种思维-非思维数据比例及三种训练计划的基准。结果揭示两种模式间存在非对称相互作用：提高非思维监督比例会降低思维模式的准确率。我们进一步表明，不同训练计划可调节该权衡，且最优计划取决于数据比例。最后，我们量化了非思维与思维模式监督间的负相关，凸显两种模式间的固有张力。这些发现为设计有效的TMF训练设置提供了实用指导，所有代码与数据已发布以支持进一步研究，地址为：Fusion Bench。
英文摘要
Thinking Mode Fusion (TMF) enables large language models to support both concise responses and long-form reasoning by unifying a non-thinking mode and a thinking mode within a single model. However, its training dynamics, including the \emph{data ratio} and \emph{training schedule} between the two modes, remain underexplored. In this work, we present a systematic study of TMF by analyzing the effects of the training schedule and data ratio between thinking and non-thinking modes. Focusing on mathematical problem solving, we construct a benchmark with multiple thinking-to-non-thinking data ratios and three training schedules. Our results reveal an asymmetric interaction between the two modes: increasing the ratio of non-thinking supervision reduces the accuracy of the thinking mode. We further show that different training schedules modulate this trade-off and that the optimal schedule depends on the data ratio. Finally, we quantify a negative correlation between non-thinking and thinking mode supervision, highlighting an inherent tension between these two modes. These findings provide practical guidance for designing effective TMF training settings. All code and data are released to support further research at: \href{https://github.com/caocongfeng/Fusion-Bench.git}{\textbf{Fusion Bench}}.
CommentsACL SRW 2026
DOI：10.18653/v1/2026.acl-srw.64
