智能体思考得更深吗？顺序规划中层间动力学的机制研究
Do Agents Think Deeper? A Mechanistic Investigation of Layer-Wise Dynamics in Sequential Planning
AI总结：
通过残差流探针、因果层跳跃干预和有效深度测量，研究了大型语言模型在自主智能体任务（多轮规划、工具使用、迭代状态更新）中层间动态的差异，发现智能体推理表现出与静态任务不同的深度分布，随着轨迹展开，模型逐步招募更多更深层，且后期出现更强的长距离层间依赖，同时残差更新从稳定特征积累转向重复校准，有效深度分析揭示了语义方向形成较早而深层对稳定最终输出仍必要的构建-精炼差距。
摘要：
最近的机制研究表明，大型语言模型（LLMs）在标准单轮任务中可能未能高效利用其深度。在自主智能体设置中，模型必须执行多轮规划、工具使用和迭代状态更新，这种情况是否仍然成立尚不清楚。我们通过系统性地对三个领域（深度研究、代码生成和表格处理）的完整用户-智能体轨迹进行逐层分析来研究这一问题。使用残差流探针、因果层跳过干预和有效深度测量，我们表明智能体推理表现出与静态任务不同的深度分布。随着轨迹展开，模型逐步招募更多和更深的层，在后期出现更强的长距离层间依赖。同时，残差更新变得越来越以校正为主导，表明从稳定的特征积累转向重复校准。有效深度分析进一步揭示了一个显著的构建-精炼差距：语义方向通常形成较早，而深层对于稳定最终输出仍然必要。在不同模型家族中，这一差距在Qwen和Minimax中显著，而GLM则表现出更依赖领域的深度分配模式。这些结果提供了机制证据，表明自主LLM智能体随着推理复杂性的增长自适应地分配深度。
英文摘要：
Recent mechanistic studies suggest that large language models (LLMs) may utilize their depth inefficiently in standard single-turn tasks. Whether this still holds in autonomous agent settings, where models must perform multi-turn planning, tool use, and iterative state updates, remains unclear. We study this question through a systematic layer-wise analysis of complete user-agent trajectories spanning three domains: Deep Research, Code Generation, and Tabular Processing. Using residual stream probes, causal layer-skipping interventions, and effective-depth measurements, we show that agentic reasoning exhibits a distinct depth profile from static tasks. As trajectories unfold, models progressively recruit more and deeper layers, with stronger long-range inter-layer dependencies emerging in later turns. At the same time, residual updates become increasingly correction-dominant, indicating a shift from stable feature accumulation toward repeated recalibration. Effective-depth analysis further reveals a substantial construction-refinement gap: semantic direction often forms relatively early, while deep layers remain necessary for stabilizing final outputs. Across model families, this gap is pronounced in Qwen and Minimax, whereas GLM shows a more domain-dependent depth allocation pattern. These results provide mechanistic evidence that autonomous LLM agents allocate depth adaptively as reasoning complexity grows.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
