System 1&2 Synergy via Dynamic Model Interpolation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Yang, Chenxu, Si, Qingyi, Tian, Chong, Liu, Xiyu, Yao, Dingyu, Qin, Chuanyu, Lin, Zheng, Wang, Weiping, Wang, Jiaqi
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908796351152128
author Yang, Chenxu
Si, Qingyi
Tian, Chong
Liu, Xiyu
Yao, Dingyu
Qin, Chuanyu
Lin, Zheng
Wang, Weiping
Wang, Jiaqi
author_facet Yang, Chenxu
Si, Qingyi
Tian, Chong
Liu, Xiyu
Yao, Dingyu
Qin, Chuanyu
Lin, Zheng
Wang, Weiping
Wang, Jiaqi
contents Training a unified language model that adapts between intuitive System 1 and deliberative System 2 remains challenging due to interference between their cognitive modes. Recent studies have thus pursued making System 2 models more efficient. However, these approaches focused on output control, limiting what models produce. We argue that this paradigm is misaligned: output length is merely a symptom of the model's cognitive configuration, not the root cause. In this work, we shift the focus to capability control, which modulates \textit{how models think} rather than \textit{what they produce}. To realize this, we leverage existing Instruct and Thinking checkpoints through dynamic parameter interpolation, without additional training. Our pilot study establishes that linear interpolation yields a convex, monotonic Pareto frontier, underpinned by representation continuity and structural connectivity. Building on this, we propose \textbf{DAMI} (\textbf{D}yn\textbf{A}mic \textbf{M}odel \textbf{I}nterpolation), a framework that estimates a query-specific Reasoning Intensity $λ(q)$ to configure cognitive depth. For training-based estimation, we develop a preference learning method encoding accuracy and efficiency criteria. For zero-shot deployment, we introduce a confidence-based method leveraging inter-model cognitive discrepancy. Experiments on five mathematical reasoning benchmarks demonstrate that DAMI achieves higher accuracy than the Thinking model while remaining efficient, effectively combining the efficiency of System 1 with the reasoning depth of System 2.
format Preprint
id arxiv_https___arxiv_org_abs_2601_21414
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle System 1&2 Synergy via Dynamic Model Interpolation
Yang, Chenxu
Si, Qingyi
Tian, Chong
Liu, Xiyu
Yao, Dingyu
Qin, Chuanyu
Lin, Zheng
Wang, Weiping
Wang, Jiaqi
Artificial Intelligence
Computation and Language
Training a unified language model that adapts between intuitive System 1 and deliberative System 2 remains challenging due to interference between their cognitive modes. Recent studies have thus pursued making System 2 models more efficient. However, these approaches focused on output control, limiting what models produce. We argue that this paradigm is misaligned: output length is merely a symptom of the model's cognitive configuration, not the root cause. In this work, we shift the focus to capability control, which modulates \textit{how models think} rather than \textit{what they produce}. To realize this, we leverage existing Instruct and Thinking checkpoints through dynamic parameter interpolation, without additional training. Our pilot study establishes that linear interpolation yields a convex, monotonic Pareto frontier, underpinned by representation continuity and structural connectivity. Building on this, we propose \textbf{DAMI} (\textbf{D}yn\textbf{A}mic \textbf{M}odel \textbf{I}nterpolation), a framework that estimates a query-specific Reasoning Intensity $λ(q)$ to configure cognitive depth. For training-based estimation, we develop a preference learning method encoding accuracy and efficiency criteria. For zero-shot deployment, we introduce a confidence-based method leveraging inter-model cognitive discrepancy. Experiments on five mathematical reasoning benchmarks demonstrate that DAMI achieves higher accuracy than the Thinking model while remaining efficient, effectively combining the efficiency of System 1 with the reasoning depth of System 2.
title System 1&2 Synergy via Dynamic Model Interpolation
topic Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2601.21414