大模型无法准确感知自身任务耗时，实验发现其预估时间普遍高估4-7倍。
Can LLMs Perceive Time? An Empirical Investigation
- 通过68项任务测试，模型预估时间比实际长4-7倍
- 在复杂任务排序中表现仅如随机猜测，准确率低至18%
- 即使多步代理场景也持续出错，误差达5-10倍，不适合时间敏感应用
大型语言模型无法估计自身任务耗时。我们通过四项实验，在68个任务和四个模型家族上研究这一局限。任务前预测的时间比实际耗时高出4至7倍（p < 0.001），模型将本应在秒级完成的任务误判为人类级别的分钟级。相对顺序判断表现同样不佳：在设计用于暴露启发式依赖的任务对中，模型得分处于或低于随机水平（GPT-5在反直觉任务对上仅18%，p = 0.033），当复杂度标签误导时系统性失败。事后回忆与真实时间完全脱节——预测值与实际值相差一个数量级。这种错误在多步代理设置中依然存在，误差达5-10倍。尽管模型在训练中具备关于持续时间的命题知识，但缺乏对自身推理时间的体验性基础，这对代理调度、规划及时间敏感场景具有实际影响。
原文摘要 · Abstract (English)
Large language models cannot estimate how long their own tasks take. We investigate this limitation through four experiments across 68 tasks and four model families. Pre-task estimates overshoot actual duration by 4--7$\times$ ($p < 0.001$), with models predicting human-scale minutes for tasks completing in seconds. Relative ordering fares no better: on task pairs designed to expose heuristic reliance, models score at or below chance (GPT-5: 18\% on counter-intuitive pairs, $p = 0.033$), systematically failing when complexity labels mislead. Post-hoc recall is disconnected from reality -- estimates diverge from actuals by an order of magnitude in either direction. These failures persist in multi-step agentic settings, with errors of 5--10$\times$. The models possess propositional knowledge about duration from training but lack experiential grounding in their own inference time, with practical implications for agent scheduling, planning and time-critical scenarios.
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。