arXiv：cs.LG· Zelin Zhao (Georgia Institute of Technology), Xinyu Guo (Georgia Institute of Technology), Jingyuan Zhang (Georgia Institute of Technology), Yuxuan Zhang (Etude AI), Yongxin Chen (Georgia Institute of Technology)·· 3 小时前AI 评分43
把 LLM 当作智能体运行要花多少算力？LAM 资源理论给出答案
Harnessing LLMs as Agents: What Does It Cost?
AI 导读
研究者提出 Language Model Agent Machine（LAM），一种固定底层语义模型、显式计入 harness 层资源开销的资源受限抽象，用于量化语言模型智能体的计算成本。
正文
Abstract:Language-model agents increasingly rely on harnesses that manage bounded context, persistent memory, tools, verification, and repeated execution, yet existing notions of model capability do not quantify the computational resources these mechanisms consume. We introduce the Language Model Agent Machine (LAM), a resource-bounded abstraction that fixes the underlying semantic model while explicitly charging harness-level resources. We establish four classes of results. Communication: LAM execution is instancewise equivalent to red--blue pebbling under simultaneous call--transfer budgets, transferring classical I/O lower bounds to context--memory traffic. Access: memory interfaces induce asymptotic separations, including a $\Theta(n)$ gap between random and non-speculative sequential access on pointer chasing. Recomputation: bit-reversal DAGs require $\Theta(n^2/(C+S)+n)$ model calls with context capacity $C$ and persistent-memory capacity $S$, quantifying when stored intermediate state avoids repeated semantic computation. Reliability: we derive tight stage-local sampling bounds, exact imperfect-verification costs, and a Young--Daly-type checkpoint law with a closed-form optimal verification interval. Controlled and held-out experiments on GPT-6 Astra test communication and reliability predictions, including checkpoint optima, policy selection under programmatic checking, and tradeoffs among call granularity, logical input traffic, and reliability on chained MATH tasks. Together, these results provide a resource theory for the computational cost of language-model agent harnesses.
| Subjects: | Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.02488 [cs.LG] |
| (or arXiv:2610.02488v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.02488 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Zelin Zhao [view email]
[v1]
Thu, 1 Oct 2026 21:05:49 UTC (1,820 KB)
来源：arXiv：cs.LG · arxiv.org