Large language models (LLMs) are increasingly deployed as autonomous agents for multi-turn decision-making tasks. However, current agents typically rely on fixed cognitive patterns: non-thinking models generate immediate responses, while thinking models engage in deep reasoning uniformly. This rigidity is inefficient for long-horizon tasks, where cognitive demands vary significantly from step to step, with some requiring strategic planning and others only routine execution. In this paper, we introduce CogRouter, a framework that trains agents to dynamically adapt cognitive depth at each step. Grounded in ACT-R theory, we design four hierarchical cognitive levels ranging from instinctive responses to strategic planning. Our two-stage training approach includes Cognition-aware Supervised Fine-tuning (CoSFT) to instill stable level-specific patterns, and Cognition-aware Policy Optimization (CoPO) for step-level credit assignment via confidence-aware advantage reweighting. The key insight is that appropriate cognitive depth should maximize the confidence of the resulting action. Experiments on ALFWorld and ScienceWorld demonstrate that CogRouter achieves state-of-the-art performance with superior efficiency. With Qwen2.5-7B, it reaches an 82.3% success rate, outperforming GPT-4o (+40.3%), OpenAI-o3 (+18.3%), and GRPO (+14.0%), while using 62% fewer tokens.
@article{arxiv.2602.12662,
title = {Think Fast and Slow: Step-Level Cognitive Depth Adaptation for LLM Agents},
author = {Ruihan Yang and Fanghua Ye and Xiang We and Ruoqing Zhao and Kang Luo and Xinbo Xu and Bo Zhao and Ruotian Ma and Shanyi Wang and Zhaopeng Tu and Xiaolong Li and Deqing Yang and Linus},
journal= {arXiv preprint arXiv:2602.12662},
year = {2026}
}