大模型在不监督中间步骤时，只能隐式规划最多7步，超出后无法执行。
The Depth Ceiling: On the Limits of Large Language Models in Discovering Latent Planning
- 用图路径寻找任务测试模型在单次前向传播中隐式规划的能力
- 最大可发现8步策略，但仅能执行至7步，存在发现与执行的断层
- 揭示了大模型隐式规划能力的天花板，支持思维链监控必要性
链式思维（CoT）监控的有效性依赖于大模型无法在其潜在表征中有效推理。然而，关于此类潜在推理的局限性仍知之甚少。我们通过研究模型在无中间步骤监督下，能否在单次前向传播中隐式发现并执行多步规划策略，来测试这些限制。利用精确控制所需潜在规划步骤数的图路径寻找任务，我们发现：从零训练的小型Transformer可发现最多三步的策略；微调后的GPT-4o和Qwen3-32B可达五步；而GPT-5.4在少量提示下可达到七步。尽管训练期间模型能学习的最大潜在这一步数为五，但发现的策略可在测试时泛化至八步。这表明，在仅以最终答案监督的情况下，模型发现潜在策略的能力与其执行该策略的能力之间存在分离。若类似限制普遍存在，则需要显式教学或外化那些需要多步协调的潜在规划策略，从而支持链式思维监控。
原文摘要 · Abstract (English)
The viability of chain-of-thought (CoT) monitoring hinges on models being unable to reason effectively in their latent representations. Yet little is known about the limits of such latent reasoning in LLMs. We test these limits by studying whether models can discover multi-step planning strategies without supervision on intermediate steps and execute them latently, within a single forward pass. Using graph path-finding tasks that precisely control the number of required latent planning steps, we uncover a striking limitation unresolved by massive scaling: tiny transformers trained from scratch discover strategies requiring up to three latent steps, fine-tuned GPT-4o and Qwen3-32B reach five, and GPT-5.4 attains seven under few-shot prompting. Although the maximum latent planning depth models can learn during training is five, the discovered strategy generalizes up to eight latent steps at test-time. This reveals a dissociation between the ability to discover a latent strategy under final-answer supervision alone and the ability to execute it once discovered. If similar limits hold more broadly, strategies requiring multiple coordinated latent planning steps may need to be explicitly taught or externalized, lending credence to CoT monitoring.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。