SAT: Balancing Reasoning Accuracy and Efficiency with Stepwise Adaptive Thinking
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866913018381598720 |
|---|---|
| author | Huang, Weiyang Bai, Xuefeng Chen, Kehai Chen, Xinyang Chen, Yibin Guan, Weili Zhang, Min |
| author_facet | Huang, Weiyang Bai, Xuefeng Chen, Kehai Chen, Xinyang Chen, Yibin Guan, Weili Zhang, Min |
| contents | Large Reasoning Models (LRMs) have revolutionized complex problem-solving, yet they exhibit a pervasive "overthinking", generating unnecessarily long reasoning chains. While current solutions improve token efficiency, they often sacrifice fine-grained control or risk disrupting the logical integrity of the reasoning process. To address this, we introduce Stepwise Adaptive Thinking (SAT), a framework that performs step-level, difficulty-aware pruning while preserving the core reasoning structure. SAT formulates reasoning as a Finite-State Machine (FSM) with distinct thinking modes (Slow, Normal, Fast, Skip). It navigates these states dynamically using a lightweight Process Reward Model (PRM), compressing easy steps while preserving depth for hard ones. Experiments across 9 LRMs and 7 benchmarks show that SAT achieves up to 40% reduction in reasoning tokens while generally maintaining or improving accuracy. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_07922 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | SAT: Balancing Reasoning Accuracy and Efficiency with Stepwise Adaptive Thinking Huang, Weiyang Bai, Xuefeng Chen, Kehai Chen, Xinyang Chen, Yibin Guan, Weili Zhang, Min Artificial Intelligence Computation and Language Large Reasoning Models (LRMs) have revolutionized complex problem-solving, yet they exhibit a pervasive "overthinking", generating unnecessarily long reasoning chains. While current solutions improve token efficiency, they often sacrifice fine-grained control or risk disrupting the logical integrity of the reasoning process. To address this, we introduce Stepwise Adaptive Thinking (SAT), a framework that performs step-level, difficulty-aware pruning while preserving the core reasoning structure. SAT formulates reasoning as a Finite-State Machine (FSM) with distinct thinking modes (Slow, Normal, Fast, Skip). It navigates these states dynamically using a lightweight Process Reward Model (PRM), compressing easy steps while preserving depth for hard ones. Experiments across 9 LRMs and 7 benchmarks show that SAT achieves up to 40% reduction in reasoning tokens while generally maintaining or improving accuracy. |
| title | SAT: Balancing Reasoning Accuracy and Efficiency with Stepwise Adaptive Thinking |
| topic | Artificial Intelligence Computation and Language |
| url | https://arxiv.org/abs/2604.07922 |